PRESENTERS

Department of Chemistry, Biochemistry & Physics  

Towards the Total Synthesis of Natural Product Scrophuloside A from Neopicrorhiza scrophulariiflora

Natural products are often used in medicinal chemistry, with natural product inspired compounds comprising up to 50% of FDA approved drugs over the last thirty years and 5% directly derived from sources. The Scrophuloside family of natural products, which are phenyl glycosides extracted from the rhizome of the plant Neopicrorhiza scrophulariiflora with far less than 0.01% isolation yield, have been reported to show moderate cytotoxic activity on P-388 murine leukemia cells and on strains of malaria. Although showing promising biological activity, there is yet to be a reported synthesis of these compounds. Described herein is the first total synthesis of Scrophuloside A in four steps, employing different strategies such as the Koenigs–Knorr reaction and its alternatives, regioselective enzymatic hydrolysis for selective deacetylation, and a dimethyltin dichloride catalyzed selective esterification of primary alcohols.The Scrophuloside family of natural products, which are phenyl glycosides extracted from the rhizome of the plant Neopicrorhiza scrophulariiflora with far less than 0.01% isolation yield, have been reported to show moderate cytotoxic activity on P-388 murine leukemia cells and on strains of malaria. Although showing promising biological activity, there is yet to be a reported synthesis of these compounds. Described herein is the first total synthesis of Scrophuloside A in four steps, employing different strategies such as the Koenigs–Knorr reaction and its alternatives, regioselective enzymatic hydrolysis for selective deacetylation, and a dimethyltin dichloride catalyzed selective esterification of primary alcohols..

RESEARCHERS: Tiffany Zhang,Valley Christian High School '26, Yeonho Noh, Mountain View High School '26
ADVISOR: Njoo Lab Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry
KEYWORDS: Organic Synthesis | Natural Products | Total Synthesis

Department of Computer Science & Engineering

Analyses of the Effect of Pulmonary Vessel Diameter on Model Focus in Image Segmentation

RESEARCHERS: Abhay Raghavendra, Milpitas High School '26; Marco Bianco, Washington High School '28; Ponkarthikeyan Saravanan, Lynbrook High School '26; Rihan Kachare, The Quarry Lane School '26

ADVISOR: Viktoriia Liu Lab, Chemistry & Computer Neurobiology & Explainable AI & Augmented Reality
KEYWORDS: Explainable AI (XAI) | Medical Imaging | Convolutional Neural Networks (CNN) | Layer-Wise Relevance Propagation (LRP) | Model Interpretability

Department of Computer Science & Engineering

Advancing Environmental Mapping and Forest Health Assessments with Drone Imaging

With the increasing concern for environmental conservation, there is a growing need for efficient methods of environmental mapping and forest health assessments. However, traditional methods employed by the U.S Forest Health Monitoring have faced controversy due to limited spatial resolution and integration of modern technologies. This research paper explores the application of machine learning algorithms in autonomous drones to conduct forest health assessments. Autonomous drones have the ability to collect timely, up-to-date data, which offers enhanced accuracy. This study focuses on training Deep Learning (DL) models to classify different environmental features based on aerial imagery captured by drones. To achieve accurate and efficient data collection, we will utilize Red-Green-Blue imaging and Convolutional Neural Networks (CNN) with the appropriate evaluation metrics, such as the Normalized Difference Vegetation Index (NDVI), Leaf Area Index (LAI), and foliage color, to create tree classes and identify forest health indicators. By integrating machine learning algorithms into forest health assessment, this study provides a more efficient, accurate, and up-to-date approach to monitor and evaluate the well-being of forests—supporting ongoing efforts towards environmental management and conservation.

RESEARCHERS: David Chin, California High School ‘26; Mina Iqlas, Foothill High School ‘28; Zhongshi Wang, Valley Christian High School ‘26; Aditya De, American High School ‘28

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS: Autonomous Drones | Machine Learning | Forest Health Assessment | Environmental Mapping

Department of Computer Science & Engineering

A Physics-Guided Hybrid Neural Network for Predicting Diffusion Monte Carlo Energies in Ground State Stereoisomers

The energy of a molecule governs many fundamental biological and chemical phenomena—ranging from protein‑folding landscapes to therapeutic drug design. Traditionally, molecular energies have been obtained by solving the Schrödinger equation. However, due to its inherent inaccuracy for multi‑electron systems, approximations such as self‑consistent field (SCF) methods and Quantum Monte Carlo (QMC) approaches have been developed. QMC methods—and in particular Diffusion Monte Carlo (DMC)—achieve near‑exact ground‑state energies by stochastically projecting the N‑body wavefunction, making them a popular choice for high‑precision energy calculations. However, DMC scales cubically with the number of electrons, O(N³), rendering it prohibitive for larger molecules. To overcome this barrier, we develop a hybrid deep‑learning framework that combines a Graph Isomorphism Network (GIN) with a physics‑guided neural network to predict ground‑state DMC energies of stereoisomers within chemical accuracy (≤ 1 kcal/mol). The GIN extracts detailed geometric and spatial descriptors from atomic graphs, while the physics‑guided network fuses these embeddings with calculated global atomic features (e.g., polarizability tensor, dipole moment) to output a DMC energy. We will train our model on molecular structures and quantum‑mechanical properties from the QM7‑X dataset and generate reference DMC energies using QMCPACK’s DMC module to create the training targets. Finally, we will benchmark the trained model against full DMC calculations to assess its predictive accuracy and computational efficiency. 

RESEARCHERS: Rishab Ghosh, Albany High School ‘26; Jayden White, Salesian High School ‘26; Sreevatsa Prevela, Monta Vista High School ‘26

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS: Molecular Energy | Diffusion Monte Carlo | Graph Isomorphism Network | Physics-Guided Neural Network | Polarizability Tensor | Dipole Moment

Department of Chemistry, Biochemistry & Physics  

Towards a Concise Total Synthesis and Structural Verification of Sporovexin A, B and C Through A Stereocontrolled Aldol

The evolutionary competition between fungal species has yielded an abundance of small-molecule antimicrobial natural products. Natural products serve a key role in medicinal chemistry, with natural product-inspired compounds comprising up to 50% of FDA-approved drugs over the last thirty years, and an additional 5% directly derived from natural sources₍₁₎. Among them, Sporovexins A–C are p-hydroxybenzoic acid metabolites of the fungus Sporormiella vexans that were demonstrated to exhibit antibiotic properties in preliminary assays₍₂₎. Using computer modeling to overlay a putative structural parallels to p-aminobenzoic acid, a substrate of the enzyme dihydropteroate synthase (DHPS) central to the bacterial synthesis of folate, we hypothesize that the antibacterial behavior of the sporovexin family of natural products arises from its competitive inhibition of DHPS. However, despite their potential, these molecules have yet to be synthesized, with prior literature exclusively focusing on their direct isolation from Sporormiella vexans₍₂₎. Here we present the synthesis of Sporovexin A and B as well as two novel des-methyl analogs of the Sporovexin family in two synthetic steps from commercially available starting materials to probe the specific effects of these functional groups on antimicrobial activity. The inhibitory behavior of each compound was determined via Kirby-Bauer testing in bacteria cultures of Bacillus cereus, Escherichia coli, Neisseria sicca, and Staphylococcus epidermidis, providing insight into the antimicrobial properties of these Sporovexin analogs.

RESEARCHERS: Jay McChesney, Dougherty Valley High School ‘26; Elizabeth Lin, The King's Academy ‘26; Anca Stefan, Cambrian Academy ‘26; Cherine Zhou, Mountain View High School ‘26

ADVISOR: Njoo Lab Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry 

KEYWORDS: Forthcoming

Department of Computer Science & Engineering

Using Quantum Machine Learning to Improve Detection of Spinal Injuries, Defects, and Illnesses

Incorporating image identification machine learning models in the medical field has prospects for improving diagnostic accuracy and efficiency, quality of individualized treatment, and research productivity.  However, the complexity of the subject leaves many areas that have yet to be thoroughly and specifically researched, such as the classification of magnetic resonance images of spinal cords and injury detection.  This research targets osteoarthritis, the abnormal growth, damage, and gradual degeneration of connective tissues, and its pathologies include osteophytes, foraminal stenosis, other forms of vertebrae misalignment or collapse that are likewise detectable through magnetic resonance imaging.  By employing a quantum convolutional neural network and visual transformer, we optimize classical computing solutions by taking advantage of the increased complexity, compression, and precision of quantum computing.  

RESEARCHERS: Alyssa Chou, George Walton Comprehensive High School ‘26; Bashar Kabbarah, Monte Vista High School ‘25; Jason Nishio, BASIS Independent Silicon Valley ‘26; Nelson Nishio, BASIS Independent Silicon Valley Upper School ‘27; Mahika Reddy, Dougherty Valley High School ‘27; Avni Saxena, Irvington High School ‘27; Sanath Beedu, Dougherty Valley High School ‘27; Jayash Patnaik, Saint Francis High School ‘28Using Quantum Machine Learning to Improve Detection of Spinal Injuries, Defects, and Illnesses

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS: Computer Science | Machine Learning | Quantum Computing | Convolutional Neural Networks | Health Technology

Department of Computer Science & Engineering

An Explainable AI Algorithm for COVID-19 Detection Based on Layer-wise Relevance Propagation

An Explainable AI Algorithm for COVID-19 Detection Based on Layer-wise Relevance Propagation

Abstract: We employ Explainable AI (XAI) to investigate the decision-making processes of a Convolutional Neural Network (CNN) built using TensorFlow, developed for classifying lung X-ray images into healthy lungs (normal), Pneumonia, and COVID-19 categories. Utilizing the XAI library, Innvestigate, while integrating Layer-Wise Relevance Propagation (LRP) and Deep Taylor Decomposition, we generate interpretive heatmaps to reveal critical regions influencing the model’s predictions. A comparative analysis of multiple LRP Rules including Alpha2, Beta1, and Z Plus, highlight differences in interpretability and precision, with methods such as LRP Z Plus and Deep Taylor Bounded offering enhanced contrast and clearer visualizations. These insights provide a detailed understanding of the model’s decision pathways, enabling the identification of potential biases and inaccuracies. Our findings not only demonstrate the utility of XAI in improving transparency and reliability in medical imaging AI, but also establish a framework for broader applications in Machine Learning where interpretability is paramount.

RESEARCHERS: Vedant Hathalia, Bellarmine College Preparatory ‘27; Tolulope Elegbede, James Logan High School ‘26; Krithi Tandyala, James Logan High School ‘26; Josh Karthikeyan, American High School ‘27

ADVISOR: Viktoriia Liu Lab, Chemistry & Computer Neurobiology & Explainable AI & Augmented Reality

KEYWORDS: Explainable AI (XAI) | Medical Imaging | Convolutional Neural Networks (CNN) | Layer-Wise Relevance Propagation (LRP) | Model Interpretability

Department of Computer Science & Engineering

Robust Mobile Application for Pneumonia Detection: Evaluating CNN Performance with Noise and Distance Variability

Pneumonia is a severe respiratory infection that inflicts the lungs and today​, it is one of the leading causes of mortality and death worldwide, particularly among children and the elderly​. Early and accurate detection of pneumonia becomes imperative for effective treatment. However, traditional diagnostic methods such as chest X-rays and clinical examinations require professional interpretation and access to medical facilities. As artificial intelligence and machine learning models become increasingly popular ​​in the healthcare industry convolutional neural networks have been used to help detect pneumonia detection using a dataset ​of chest X-ray images. Mobile applications ​using CNN-based models ​​also offer an encouraging and hopeful solution to improve medical diagnoses of pneumonia​, as some diagnostics may be challenging and difficult to identify from the naked eye. ​​By doing so, patients can receive proper treatment. However, the accuracy of these CNN models’ predictions can be influenced by a wide variety of factors, including image quality resolution, and the distance that the chest X-ray image is captured or taken before being processed by the CNN model in the mobile app. This research aims to evaluate the effect of photo capture distance of a chest X-ray image on image quality ​and the predictions made by a CNN-based​ ​pneumonia detection model in a mobile application

RESEARCHERS: John Xie, Bellarmine College Preparatory ‘27; Vihaan Udayakumar, Mission San Jose High School ‘28

ADVISOR: Viktoriia Liu Lab, Chemistry & Computer Neurobiology & Explainable AI & Augmented Reality

KEYWORDS: Machine Learning | Convolutional Neural Networks | Computer Science 

Department of Computer Science & Engineering

Hybrid Quantum-Classical Graph Generative Models To Target Alzheimer's Disease

Contemporary drug discovery and development processes require billions of dollars and lengthy amounts of time, which is why researchers are utilizing computational chemistry methods like machine learning to speed up molecular synthesis pathways. Alzheimer's disease presents a specific challenge, as treatment requires molecules that can penetrate the blood-brain barrier (BBB). There are currently no effective cures for Alzheimer's disease. Our group is currently working on developing a generative adversarial network to generate chemically stable, novel, and druglike molecules that are of the right size and fit the molecular descriptors to penetrate the BBB. Not only this, but it must also inhibit the protein amyloid beta, which accumulates in the brain and causes Alzheimers. To accomplish this we have implemented many algorithms to verify the molecule’s efficacy like a size checker and toxicity verifier. In previous work, our group developed QNetGAN v2, a model that utilizes quantum computing and generative adversarial networks (GANs) to generate chemically feasible molecules for general drug discovery with an 91.0% success rate, generating 273 out of 300 structurally valid molecules.

RESEARCHERS: Aditya Pendyala, The Quarry Lane School ‘27; Sahil Vijay, Leigh High School ‘26; Anjali Kukkamalla ‘26

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS:  Machine Learning | Quantum Computing | Drug Discovery | Computational Chemistry | Alzheimer's Disease

Department of Computer Science & Engineering

Benchmarking SNN, CNN, and RNN Combination Models to Translate ASL for Bionics Use-cases

The creation of more usable and human-friendly biological prosthetics is of vital importance to the field of disability and rehabilitation engineering. In recent years, a variety of model architectures have been developed to closely mimic the structure of sensory organs, particularly in processing and gathering visual information. Comparing these models provides valuable insight into how prosthetics can process visual information. We hypothesized that hybrid architectures—specifically those combining spiking neural networks and convolutional layers (SNN-CNN)—would outperform both standalone and other hybrid configurations by effectively integrating spatial and temporal features. This project benchmarks the variations of the SNN, CNN, and LSTM (RNN) models on American Sign Language video data provided by the Google Isolated Sign Language Recognition dataset. By processing the video data of the first 250 words taught to infants in sign language, we mimic the biological organization of visual data. Through the testing of these models, we determine the SNN-CNN combined model to be the most accurate in understanding complex processed sequential data, followed by the LSTM-CNN, SNN-LSTM, SNN, and LSTM models respectively.

RESEARCHERS: Jaiveer Gill, Bellarmine College Preparatory ‘26; Francesca Yang, BASIS Independent Silicon Valley ‘27; Pradyun Sethupathi, James Logan High School ‘26

ADVISOR: Subramaniam Lab, Machine Learning/Data Science

KEYWORDS: Machine Learning | Prosthetics | Computer Science

Department of Computer Science & Engineering

Implementing eXplainable AI (XAI) Techniques with the Captum Library for Interpretable CNNs in Disease Detection

Explainable AI (XAI) is a subset of artificial intelligence (AI) that focuses on creating models and systems that can provide human-understandable explanations for their decisions and actions. Explainable AI models are gaining importance due to the focus on human-centered design principles, taking into account the needs, expectations, and cognitive abilities of users. Unlike other deep learning networks, which are often “black boxes”, explainable AI models are designed to be transparent, meaning they provide visibility into how they arrive at their predictions or decisions. Thus, XAI models are designed to provide interpretable and understandable explanations for their decisions and actions, allowing people to understand how and why the AI system makes certain predictions or decisions.

 XAI models can provide valuable assistance to radiologists and other healthcare professionals in interpreting X-ray images in several ways. First, XAI models can provide explanations for their predictions, allowing radiologists to understand the reasoning behind the model's decision. For example, an XAI model could highlight areas of the X-ray image that have most affected the prediction, or provide textual or visual explanations that describe features or patterns that contributed to the prediction. This can help radiologists verify model results, gain insights into the model's decision-making process, and build confidence in the reliability of the model. Second, such improved models can help identify potential errors or biases in the X-ray image interpretation process. For example, if the model's explanation shows that the prediction was based on a small area of the image, the radiologist can double-check that region for any artifacts or misinterpretations. This can serve as a valuable error-checking mechanism and improve the overall accuracy of X-ray image interpretations. Finally, XAI models can enhance trust and transparency in the AI-assisted interpretation of X-ray images. XAI models can provide interpretable explanations that help radiologists explain and communicate the basis for their diagnoses, increasing transparency and confidence in the decision-making process. In this research, we apply explainable AI to X-ray images of the lungs in order to diagnose lung infections such as pneumonia and abnormal buildup of fluid in the lungs. 

RESEARCHERS: Aayush Tarhalkar, Irvington High School ‘26; Adhvaith Vinothkumar, Round Rock High School ‘26; Advi Wehzan, Leland High School ‘26; Anirudh Ayyadevara, Dougherty Valley High School ‘25; Roshan Saxena, Monta Vista High School ‘27; Trisha Sallakonda, Emerald High School ‘28

ADVISOR: Viktoriia Liu Lab, Chemistry & Computer Neurobiology & Explainable AI & Augmented Reality

KEYWORDS: Forthcoming

Department of Computer Science & Engineering

Machine Learning in Electron Microscopy Image Analysis: Particles Identification 

In the field of electron microscopy, elements of artificial intelligence and machine learning (ML) have become increasingly prevalent. The numerous tasks of using a scanning electron microscope (SEM) are very inefficient due to their time consuming, repetitive, and manual nature. Yet, these processes require significant degrees of accuracy. We intend to focus on using ChatGPT generated ML modified to analyze self-assembly in nanoparticles in SEM images. Using both spherical and cuboidal nanoparticles in both ordered and disordered datasets, we have begun by binning the different nanoparticles based on frequency. SEM image datasets found online will be used for training, validating, and testing.

RESEARCHERS: Sophia Hale, Saint Francis High School ‘25; Anderson Lu, BASIS Independent Silicon Valley ‘28; Anannya Kaur, Valley Christian Schools ‘26

ADVISOR: Starostina Lab, Materials Science

KEYWORDS: Deep Learning | Scanning Electron Microscopy | Nanoparticle Detection | Image Analysis

Department of Computer Science & Engineering

Towards Predictive Maintenance:  Tabor Factor Determination as a Function of Strengthening Mechanisms in Copper

Determining the Tabor factor in relation to microstructure and composition could pave the way for the creation and development of an inexpensive, non-destructive method for predicting the tensile properties of bulk materials using localized hardness measurements. This advancement is especially valuable for improving current preventative maintenance procedures and facilitating the upscaling of research and development in industrial settings.To start our research, we acquired CAD models and followed machine ASTM E8 standard tensile testing (TT) procedures for our copper samples. TT was performed at Santa Clara University, and five stress-strain (SS) diagrams were created and analyzed to determine their flow stress. Additionally, grain size was measured on both sides of the sample to be correlated alongside the flow stress. The microstructures and SS data will be shared and discussed in terms of the literature searches.

RESEARCHERS: Anay Tailor, Dougherty Valley High School ‘27; Nicholas Wong, Dougherty Valley High School ‘26; Saketh Potturu, Dougherty Valley High School ‘26; Shouyo Tan, Los Altos High School ‘26, Timothy Park, Dublin High School '26; Erin Wu, Henry M. Gunn High School '26; Averi Mukhopadhyay, American High School '27

ADVISOR: Starostina Lab, Materials Science

KEYWORDS: Tabor Factor | Microstructure | Tensile Properties | Predictive Maintenance

Department of Computer Science & Engineering

Utilizing Autonomous Drones to Identify and Target Fire Hazards in Close Proximity to Power Lines in Remote Areas

In the past few years, the amount of forest fires in the United States has skyrocketed, many of they caused by downed or damaged power lines catching fire. The goal of this project is to use automated drones to analyze the area near a powerline and to determine whether it is in danger of being damaged, and if it is, the drone will send the data back to a database where local authorities will able to decide on the action they need to take. The main hazard that we will be trying to pinpoint is the vegetation around the power lines and the possible danger it may possess to the power lines. We are using a regular drone and a couple other parts to make it fly. To capture and analyze the picture, we will be taking a video of the powerlines. We will then split the video into still frames and put that through an image processor which we will train and make. Once this is all done, we will know if the powerline is in danger, and if so will notify the local authorities. In the future, we plan to begin to test our drone in parks and other places to see if it works and functions in the way we expect it to.

RESEARCHERS: Nirupama Balaji, American High School ‘28; Isha Ramakrishna, Foothill High School ‘27

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS:  Engineering | Research and Development | Machine Learning | Autonomous Drones

Department of Chemistry, Biochemistry & Physics  

Reactivity-informed Pharmacophore Editing and Biological Evaluation of Andrographolide and its A-ring Analogs

Natural products and their analogs have long served as inspiration for the exploration and development of small molecules with therapeutic significance. One such compound is andrographolide, a labdane diterpenoid extracted from the plant Andrographis paniculata, which has been extensively studied as an anticancer therapeutic. It is known to function putatively through covalent inhibition of NF-kB, a transcription factor at the crossroad of a myriad of cell signaling pathways that modulate tumor survival and metastasis. Functionalization of the C-19 hydroxyl might alter the primary mode of action from inhibition of NF-kB to the modulation of the Wnt/𝜷-catenin signaling pathway. To interrogate the structure-activity relationship of this position, we synthesized a library of andrographolide analogs by protecting the C-19 hydroxyl with large, hydrophobic silyl and trityl ethers. Inspired by the observed biological trends amongst the library, we sought to further interrogate more complex A-ring oxy-functionalization. Among several targets for A-ring functionalization, we were intrigued by early isolation of an A-ring oxetane analog that is a biosynthetic byproduct isolated from A. paniculata (Jantan et al., Phytochemistry 1994), whose synthetic preparation and biological properties are not well described. After investigation of several synthetic route candidates, we identified an efficient route to access the A-ring oxetane with three chromatographic purifications. En route, we describe mechanistic insight into A-ring reactivity of andrographolide and its analogs.

RESEARCHERS: Rushika Raval, Irvington High School ‘26

ADVISOR: Njoo Lab Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry 

KEYWORDS: Andrographolide | Wnt/𝜷-catenin | NF-kB | Natural product analogs | Oxetane

Department of Computer Science & Engineering

Interfacial Free Energy of Copper: Experimental and Computational Approaches

Knowledge of interfacial free energy is of major importance for theoretical, computational, and practical applications such as developing bio-compatible materials, colloids, optical devices and metal-matrix composites. There is no universally reliable and convenient method to perform experimental determination of solid-solid interfaces. There are computational methods, and direct and indirect experimental methods. In this study we will overview the major methods for solid copper. We will be introducing atomic force microscopy (AFM) as a new method. AFM will be compared to zero-creep and optical interferometry methods, as well as the outcomes of computational models in terms of ease of use, cost, time, resolution of the technique, and accuracy of the results. Our results on sample preparation for dihedral measurements with the AFM will be shared and discussed.

RESEARCHERS: Phinna Yin, Washington High School ‘26; Siqi Feng, Basis Independent Silicon Valley ‘28; Larry Xie, Milpitas High School ‘27; Darvas Gao, Washington High School ‘25; Saahithi Srikanth, Monta Vista High School ‘27; Aansh Chopra, Washington High School ‘26; Seoyeon Kim, Valley Christian High School ‘26; Gabriela Saree Formanek, Notre Dame Belmont High School ‘26

ADVISOR: Starostina Lab, Materials Science

KEYWORDS: Interfacial free energy | Solid-solid interface | Interfacial energy measurement | Interfacial tension | Surface energy | Copper interface

Department of Computer Science & Engineering

Evaluation of Mechanical Properties of 3D-printed PLA in Two Printing Orientation and between Two printers 

Additive manufacturing is an alternative technology to traditional manufacturing applicable to a range of fields from structural and civil engineering to dental and precision manufacturing. Recent studies show that the mechanical properties of 3D printed parts depend on printing parameters, color and brand of the feedstock and brand of the printer. This study is focused on comparing the mechanical properties evaluated by ASTM tensile test between two printing orientations (vertical and horizontal), and between two 3D printers (Bambu and Prusa).The results will be compared to the literature and discussed. 

RESEARCHERS: Divya Eashwer, American High School ‘26; Hanming Zhao, Independence High School ‘25

ADVISOR: Starostina Lab, Materials Science

KEYWORDS: Material Science | 3D printing | Material Characterization 

Department of Computer Science & Engineering

The Development of an AI-Enhanced Autonomous Drone for Pollution Monitoring and Emission Predicting

In the last 60 years, the rate of atmospheric carbon dioxide has increased around 100 times compared to the last ice age. (climate.gov) As the levels of carbon dioxide (CO2), methane, and other greenhouse gasses increase, the impact on Earth’s ecosystem is detrimental to climate and biodiversity. Research has been conducted on this topic, and attempts have been made to reduce pollution levels and greenhouse gas emissions. Such attempts include research to improve electric cars to promote less emissions from cars along with wind turbines and other forms of clean energy.

Our research involves implementing machine learning algorithms into an autonomous drone to reduce pollution levels and greenhouse gas emissions. We plan to train our algorithm with numerical data consisting of the concentration of nitrogen, CO2, and methane molecules for atmospheric conditions. After training, the drone will collect testing data from the atmosphere and predict the concentrations of these greenhouse gasses. Based on the concentration in a given area, the drone will adjust its navigation toward locations of high concentration to determine possible emission sources. With these findings, we can implement clean energy sources to reduce emissions and better our environment. 

The aim is to use the drone’s predictions to determine the areas that need the most assistance in reducing emissions. The drone will continuously record the pollution levels for different areas and give highly updated readings based on the greenhouse gas concentrations. Currently, manual pollution monitoring is costly, ineffective, and inefficient. By constantly giving accurate readings with the concentrations, our autonomous drone limits the need for manual testing, which saves time and allows us to find solutions for highly polluted areas. 

RESEARCHERS: Christon Rex, California High School ‘28; Haramrit Bal, Mountain House High School ‘27; Xiangtuo Cui, Basis Independent Silicon Valley ‘26; Rohan Devnani, Emerald High School ‘27; Trung Duong, Evergreen Valley High School ‘26; Komal Jasuja, Weston High School ‘27

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS:  Pollution, Autonomous Drones, Artificial Intelligence, Emission Reduction, Pollution Monitoring

Department of Chemistry, Biochemistry & Physics  

Synthesis and anticancer properties of triacetate and acetonide analogs of Proscillaridin A

Several cardiac glycosides, including digoxin, digitoxin, and proscillaridin A, have been originally identified as cardiomyocyte modulators and are currently being investigated for their anti-cancer properties. These cardiac glycosides are generally classified into cardenolides and bufadienolides, which bear butenolide and pyrone D-ring functionality, respectively, and have exhibited remarkable in vitro toxicity in various cancerous cell lines. As simple modifications on steroidal small molecules have demonstrated success in augmenting bioavailability or enhancing downstream biological activities, we sought to prepare synthetic prodrugs of proscillaridin A, a bufadienolide isolated from the genus Scilla. We synthesized two novel analogs of proscillaridin A bearing acetate esters or dimethyl ketals to investigate how strategies of ketalization or acetylation of the A-ring allylic glycoside might alter its anti-cancer properties. The antiproliferative activity of these compounds was evaluated alongside proscillaridin A and two model cardiac glycosides—digoxin and digitoxin—across several colorectal and liver cancer cell lines. Through a diverse panel of cell viability and cytotoxicity experiments, reporter assays and apoptosis and cell cycle analysis by flow cytometry, we find that ketalization of the glycan of proscillaridin A provides similar, and in some cases enhanced, in vitro potency. This study establishes the foundation for current and further in vitro and in vivo evaluation.

RESEARCHERS: Hasini Menta, California High School '25; Shreya Somani,  Lynbrook High School '26

ADVISOR: Njoo Lab Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry 

KEYWORDS: Forthcoming

Department of Computer Science & Engineering

Quantum Computing Error Mitigation

Quantum computing has the potential to significantly improve computational tasks. Unfortunately errors due to outside interactions cause the data to be noisy. Our group is focusing on finding a way to decrease these errors by reverting the data back to what it was before the noise. To simulate realistic errors, we used a bosonic bath model on surface code. We use both Convolutional Neural Networks (CNNs), for preprocessing and feature extracting, and Graph Neural Networks (GNNs), for predicting and correcting errors in the surface codes. By training and evaluating the CNNs and the GNNs, we can increase performance in quantum computing. These changes will allow for major developments in the field and will be a significant contribution to making an accurate quantum structure.

RESEARCHERS: Aryan Das, Amador Valley High School '25; Satvik Dronavalli, Independence High School '26; Ayushman Bisht, Bellarmine College Preparatory '28

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS: Quantum Computing | Quantum Error Correction | Machine Learning | Convolutional Neural Networks | Graph Neural Networks

Department of Computer Science & Engineering

Theoretical study of quantum dots in noise-resistant quantum key distribution systems

This research project is meant to analyze ways quantum dots can be better integrated into Quantum Key Distribution (QKD) networks to counter interference impacts such as noise and detect the presence of eavesdroppers. In opposition to typical private key encryption, QKD encryption uses quantum effects to guarantee the security of a private key over an open channel. However, currently, the presence and magnitude of noise, or interference from the environment such as heat, electronic noise, etc, degrades the security of QKD systems and makes it easier for eavesdroppers to potentially access the key. Through the results gathered through advanced simulations, which ran on a high-performance computing cluster simulating QKD systems, new insight into how unique quantum dot properties can improve noise resistance emerged from the data. These findings suggested potential developments in security and privacy, which sprouted from the implementation of newer quantum communication protocols and channels. In addition, this research could lead to more robust QKD systems that can be utilized in sectors where privacy is important such as telecommunications and national security.

RESEARCHERS: Aditya Das, American High School '26; Sophia Ren, The Madeira School '26; Elaine Huang, Harker Upper School '28

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS: QKD | Quantum Dots | Noise Model | BB84 protocol | Computer Science | Quantum Computing | Quantum Physics

Department of Chemistry, Biochemistry & Physics  

Synthesis and anticancer properties of triacetate and acetonide analogs of Proscillaridin A

Several cardiac glycosides, including digoxin, digitoxin, and proscillaridin A, have been originally identified as cardiomyocyte modulators and are currently being investigated for their anti-cancer properties. These cardiac glycosides are generally classified into cardenolides and bufadienolides, which bear butenolide and pyrone D-ring functionality, respectively, and have exhibited remarkable in vitro toxicity in various cancerous cell lines. As simple modifications on steroidal small molecules have demonstrated success in augmenting bioavailability or enhancing downstream biological activities, we sought to prepare synthetic prodrugs of proscillaridin A, a bufadienolide isolated from the genus Scilla. We synthesized two novel analogs of proscillaridin A bearing acetate esters or dimethyl ketals to investigate how strategies of ketalization or acetylation of the A-ring allylic glycoside might alter its anti-cancer properties. The antiproliferative activity of these compounds was evaluated alongside proscillaridin A and two model cardiac glycosides—digoxin and digitoxin—across several colorectal and liver cancer cell lines. Through a diverse panel of cell viability and cytotoxicity experiments, reporter assays and apoptosis and cell cycle analysis by flow cytometry, we find that ketalization of the glycan of proscillaridin A provides similar, and in some cases enhanced, in vitro potency. This study establishes the foundation for current and further in vitro and in vivo evaluation.

RESEARCHERS: Hasini Menta, California High School '25; Shreya Somani,  Lynbrook High School '26

ADVISOR: Njoo Lab Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry 

KEYWORDS: Forthcoming

Department of Computer Science & Engineering

Analyzing 3-D Ultrasonography using Deep Learning and Convolutional Neural Networks to Detect Breast Cancer

Breast Cancer is the most common cancer found in women. Many types of imaging including ultrasound, mammography, CT, MRIs, and X-Ray are used to diagnose breast cancer. Most methods of imaging that were mentioned are expensive and require radiation. Ultrasound, on the other hand, is relatively inexpensive and does not require radiation. However, ultrasonography creates low-definition depictions of the breasts compared to other methods of imaging, which causes difficulty of detecting cancer compared to the others.

Many studies have shown that computed tomography, mammography, MRI, and X-Rays can train a CNN model to diagnose cancer accurately. There are also studies that also show that 2D ultrasounds can be used to diagnose cancer accurately too. Recent studies show that 3D-TVUS is comparable to MRI as far as diagnostic accuracy is concerned, however remains cheaper, less time-consuming and more tolerable, while offering some advantages over 2D-TVUS as well [7]. However, the characteristics determined on the conventional planes of 3D ultrasound differed from those determined on the 2D ultrasound images. The diagnostic accuracy of 2D and 3D ultrasound in the ROC analysis was almost identical (area under the curve 0.846 and 0.851, respectively) [8].

Using Deep Learning and CNNs, we will create a model that detects breast cancer by analyzing 3D ultrasound images. We will compare multiple CNN architectures, including a ResNet-18 and a VGG-16 architecture. All will be for feature extraction. Our model will also include a Max Pooling Layer which shrinks the feature vector while keeping the most informative features. The remaining features will be concatenated and will be put through an output layer which will determine whether there is breast cancer present in the image.

We want to show that ultrasonography should be primarily used for diagnosis of cancer. We hope to find the best CNN architecture for detecting cancer in 3D ultrasound images taken into the fact the trade-off between computing power and accuracy.

RESEARCHERS: Andrew Lin, BASIS '27; Gautam Taneja '27 Cupertino High School

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS: Ultrasound | Ultrasonography |  Deep Learning |  Convolutional Neural Network |  CNN |  Breast Cancer |  Cancer

Department of Chemistry, Biochemistry & Physics  

Development of bifunctional small molecules for localized ER-targeting chemotherapy for breast cancer AND Synthesis and Biological Evaluation of Novel Anti-Cancer Silyl Ether SN-38 Analogs 

Historically, most medicinal chemistry campaigns have focused on a paradigm of single small molecule modalities targeting sole biological targets. Bifunctional chemical entities incorporating two compounds with discrete and distinct functionalities tethered together with a linker may provide access to higher orders of molecular logic. More recently, novel heterobifunctional small molecules called Regulated Induced Proximity Targeting Chimeras (RIPTACs), which operate through localization of a ligand-toxin conjugate to induce selective apoptosis, have demonstrated enhanced anti-proliferative effect in cancer treatment, alongside increased scalability. Estrogen receptor (ER)-positive breast cancer accounts for a majority of U.S. breast cancer cases, affecting hundreds of thousands of women annually. We hypothesized that the upregulation of ER found in these cancer cells could be leveraged to drive localization of possible RIPTAC compounds with ER ligands to selectively induce apoptosis. With this in mind, we are currently designing and synthesizing a library of bifunctional targeting ligands consisting of various ER-localizing ligands tethered to SN-38, a topoisomerase I inhibitor and effector protein toxin. To evaluate the anti-proliferative effects of our bifunctional conjugates, especially in comparison to their standalone components, we performed MTT assays on colorectal and both ER+ and ER- breast cancer cell lines and found increased anti-proliferative activity of our compound raloxifene-SN-38 in ER+ cell line compared with raloxifene, but reduced activity in ER- cell line in comparison to SN-38.

While silicon makes up over 27% of the earth’s crust, it accounts for less than 1% of patented compounds in drug discovery spaces. In previous drug discovery spaces, silicon-containing compounds have shown favorable pharmacokinetic benefits, such as increased cell penetration and metabolic stability. Camptothecin, a natural product alkaline derived from Camptotheca acuminata, has been previously investigated for its potent anti-cancer activity. Among its synthetic analogs, silane compounds such as DB-67 and karenitecin, have been found to be more potent than their predecessor. Inspired by this, we designed and synthesized three siloxy derivatives of SN-38, a camptothecin analog, with the aim of probing the effects of C-9 silylation and structural bulk on anti-proliferative activity. Our biological evaluation was conducted on a variety of breast, colorectal, and lung cancer cell lines through MTT assays. Through these studies, we discovered that our silyl ethers, in particular TBDPS-SN-38, showed comparable and sometimes greater anti-proliferative effects when evaluated alongside SN-38 and camptothecin. In addition to our MTT assays, we conducted BrdU and topoisomerase I assays to confirm the functionality of our siloxy derivatives. We are currently in the process of synthesizing C-9 siloxy ether SN-38 antibody-drug conjugates for targeting triple-negative breast cancer tumors in an effort to further explore drug localizing strategies.

RESEARCHERS: Sanika Vaidya , Lynbrook High School '26; Alyssa Chia, The King's Academy '26; Ashley Mo, The Harker School '26

ADVISOR: Njoo Lab Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry 

KEYWORDS: Forthcoming

Department of Computer Science & Engineering

Using Quantum Neural Networks (QNNs), Quantum Vision Transformers (QVT), and the Mathematical Morphological Reconstruction Algorithm (MMR) for Brain Tumor Detection

Brain tumors affect millions around the world, so detection is critical to helping doctors determine treatment. Currently, radiologists manually identify tumors through MRI (Magnetic Resonance Imaging) scans; however, this poses several limitations: it creates a heavy reliance on the experience of radiologists, has become increasingly costly and time-consuming, and is not as accessible to areas that lack the necessary resources and doctors. With the advancement of deep learning algorithms, a more accessible and efficient solution is possible. Given the existing research in classical Convolutional Neural Networks (CNNs) for tumor detection, Quantum Convolutional Neural Networks (QCNNs) and Quantum Vision Transformers (QVT) offer a promising approach to the problem. Mathematical Morphological Reconstruction (MMR), another image processing method, provides a relative metric for success in the QCNN, and is another classical alternative to CNNs. This research compares the accuracy and computational speed of the MMR, QCNN, QVT, and CNN algorithms to determine whether introducing a quantum aspect presents any noticeable advantage. To build these models, extensive datasets of MRI brain scans were collected. The MMR algorithm involved applying various techniques such as dilation, erosion, and skull stripping through OpenCV2's morphology functions. The QCNN algorithm utilizes quantum power to encode the data into a parametrized quantum circuit and apply convolutional and pooling layers. In terms of future steps, QVTs will be implemented with QCNNs for higher spatial understanding. So far, our results indicate that the MMR algorithm achieved up to 92% accuracy. These results will be compared with the accuracy of the QCNN, QVT, and CNN algorithms.

RESEARCHERS: Tiffany Liu, The Quarry Lane School '25; Riddhi Sharma, Evergreen Valley '26; Eesha Gadekarla, The Quarry Lane School '25 

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS: Quantum Computing | Brain Tumors | Machine Learning | Mathematical Morphological Reconstruction | Convolutional Neural Networks

Department of Computer Science & Engineering

Autonomous Drone-Based Detection of Invasive Plant Brassica Nigra Using Machine Learning

Invasive species, like Brassica Nigra (Black Mustard Plant), are an increasing problem in Northern California because they threaten local plants and wildlife, harming ecosystems greatly. This study focuses on creating an autonomous drone capable of detecting and tracking areas of Brassica Nigra to help limit its spread. The research question was whether drones, using cameras and artificial intelligence, could accurately detect Brassica Nigra from different angles and distances. We hypothesized that an autonomous drone could successfully identify Brassica Nigra using a machine learning model. The drone consists of a Holybro X500 V2 ARF drone frame that is flexible for all the hardware components, and is flown with a Pixhawk controller The results indicate that the drone can accurately identify Brassica Nigra and distinguish it from native plants in various locations. Additionally, the drone can correctly send locations of areas populated with Brassica Nigra using a Holybro M10 GPS. The methods included compiling an image dataset of about 100,000 photos featuring both native plants and Brassica Nigra, then flying the drone over areas with Brassica Nigra, testing the drone’s detection accuracy and location reporting. These high-accuracy results suggest that autonomous drones can be a new cost-effective tool to control the spread of invasive plant species over large areas.

RESEARCHERS: Chloe Ho, Basis Independent Silicon Valley '26; Sahiti Pantangi, Washington High School '28; Yashvee Shah, Santa Clara High School '26, Alex Liu, Basis Independent Fremont '26

ADVISOR: McMahan Lab, Quantum Computing & Computer Science 

KEYWORDS: Invasive Species | Machine Learning Model | Autonomous Drone

Department of Chemistry, Biochemistry & Physics  

Anticancer synthetic arylsulfonamides with Wnt1-modulating activity

Dysregulation of the Wnt1/β-catenin signaling pathway has been demonstrated to be a driving factor in the propagation of several human cancers. Previous studies have discovered methyl 3-{[(4-methylphenyl)sulfonyl]amino}benzoate (MSAB) as a selective inhibitor of the Wnt1/β-catenin signaling pathway, which putatively functions through direct engagement of β-catenin. To understand how changes to the identity and position of the methyl ester affect the in vitro potency of this compound in Wnt1-driven mammalian cell lines, we prepared and evaluated three analogs of MSAB with 3- and 4-substituted methyl and ethyl esters. In MTT assays, analogs with methyl esters showed significantly more activity than their ethyl ester counterparts and both 4-substituted esters exhibited significantly attenuated antiproliferative activity, with MSAB exhibiting dose-dependent activity across cancerous cell lines. Further analysis by flow cytometry reveals low-Annexin V signal, suggesting that these compounds do not function via a pro-apoptotic pathway. Additionally, through a TCF/LEF-activated luciferase reporter cell assay, we observe that the 4-substituted methyl ester analogous to MSAB exhibits slightly diminished Wnt1-inhibitory activity, while 3- and 4-substituted ethyl esters exhibit minimal Wnt1-inhibitory activity. This difference in potency with a simple ester substitution might be attributed to several factors that ultimately drive antiproliferative activity, prompting the investigation of other potential substituents to further investigate the structure-activity relationship of these compounds as Wnt1-based antiproliferative agents.

RESEARCHERS: Vihaan Sharma, Irvington High School '26;  Emilia Lee, Carlmont High School '26

ADVISOR: Njoo Lab Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry 

KEYWORDS: Forthcoming

Department of Chemistry, Biochemistry & Physics  

Synthesis, characterization, and paramagnetic properties of novel 5-substituted cytosine complexes of copper (II)

Among the pyrimidines, cytosine metal complexes and their derivatives have elicited considerable interest due to their broad potential for bioactivity, including but not limited to their antimicrobial, anticancer, and antiproliferative properties. Cytosine copper complexes have been previously reported, most recently responsible for the TET inhibitory effects of cytosine derivative Bobcat339, but the factors contributing to their properties have yet to be fully elucidated upon. Here we disclose the effects of the alteration of the carbon five substitution of cytosine on the mode of coordination and binding affinity. Samples of [Cu(Cyt)2](OTf)2 and [Cu(XCyt)2](OTf)2 (X = F, Cl, Br, I, Me) were prepared and characterized via UV-VIS, FT-IR spectroscopy, ESI-MS, Cyclic Voltametry; and 1H, 19F NMR experiments. We further report that substituted cytosine complexes (with the exception of [Cu(FCyt)2](OTf)2 are easily differentiable from their non-substituted counterparts via their dissociation in water into non-complexed constituents.

RESEARCHERS: Abhinav Chalasani, Mission San Jose High School '26; Harahng Lee, The King's Academy '25

ADVISOR: Njoo Lab Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry 

KEYWORDS: Forthcoming