PRESENTERS

Department of Computer Science & Engineering

Deep Learning-Enabled Precision Segmentation of Nanoparticles in Scanning Electron Microscopy (SEM)

Abstract Upcoming

RESEARCHERS: Christina Tian, The Harker School ‘28

ADVISOR:  Starostina Lab, Materials Science
KEYWORDS: Materials Science | Machine Learning | Artificial Intelligence | Convolutional Neural Networks | Scanning Electron Microscopy | Nanoparticles

Department of Chemistry, Biochemistry & Physics

Benchmarking Deep Learning Architectures for Early-Stage Brain Tumor Detection & Classification 

Abstract Upcoming

RESEARCHERS: Shambhavi Jyothirganesh, Lynbrook High School '29; Siddhanth Venganti, Mountain House High School '29

ADVISOR: McMahan Lab - Quantum Computing, Computer Engineering
KEYWORDS: Drone Technology | Bike Lane Safety | Infrastructure Monitoring | Hazard Detection | Sustainable Transportation

Department of Chemistry, Biochemistry & Physics

Progress Toward the Synthesis of Pentacyclic Triterpenoid Analogs  

Abstract Upcoming

RESEARCHERS: Kristen Cheng, Los Altos High School '29

ADVISOR: Njoo Lab, Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry
KEYWORDS: Organic Synthesis | Natural Products | Pentacyclic Triterpenoid | Chemical Biology 

Department of Computer Science & Engineering

Human-Aligned Auto-Grading of LLM-Generated Restaurant Recommendations Using Multi-Model Ensembles

Large language models (LLMs) offer flexible, personalized restaurant recommendations but continue to exhibit weaknesses in spatial awareness, response consistency, and niche-query accuracy. Rapid evaluation is essential for improving these systems, yet manual grading is too time-intensive for iterative development. This study presents a hybrid autograding framework designed to approximate human evaluation while reducing grading time from weeks to minutes. Initial autograding showed substantial divergence from human evaluations; thus, we employed two strategies to improve human–LLM alignment: rubric-focused prompt engineering and a hybrid approach that combines a Human Verification Layer for accuracy with an Automated LLM Ensemble to assess adherence to user criteria. These modifications increased agreement to R^2 = 0.753. We validated the framework by improving restaurant recommenders and evaluating the resulting responses using both human and automated grading, demonstrating that the framework was able to successfully quantify improvement in recommender performance while approximating human scoring. These findings suggest that hybrid human–LLM grading can serve as a reliable, time-saving proxy for restaurant recommendation evaluation while reinforcing the continued importance of human oversight in LLM-related tasks. By enabling rapid, large-scale prompt testing and model iteration, this framework accelerates the development of domain-specialized “food-expert” LLMs and provides a generalizable methodology for scalable human-aligned evaluation–an essential step toward trustworthy AI systems in real-world consumer applications.

RESEARCHERS: Abraham Yeung, American High '27; Isabel Zheng, The King's Academy '28

ADVISOR:  Qin Lab, AI & Machine Learning
KEYWORDS: Large Language Models (LLMs) | Auto-grading | Recommendation Systems | Domain-specialized AI | Human-LLM Alignment

Department of Chemistry, Biochemistry & Physics

Sublethal Antibiotic Stress Drives Mutagenesis and the Emergence of Antibiotic Resistance

Antimicrobial resistance is a growing public health threat driven by the ability of bacteria to adapt under antibiotic pressure. While resistance can arise from spontaneous mutations and regulatory changes, increasing evidence suggests that sublethal antibiotic exposure actively promotes mutagenesis. Sub-minimal inhibitory concentrations (sub-MIC) of antibiotics induce stress, enabling the emergence and spread of resistance. In this study, we investigate how sub-MIC antibiotic stress drives mutagenesis in Escherichia coli. We focus on stress-induced mutation mechanisms associated with DNA double-strand breaks (DSBs), which serve as a key source of adaptive genetic variation. Using tetracycline-sensitive (TetS) strains carrying a +1 frameshift mutation in the tetR gene, we monitor the acquisition of secondary mutations required to restore tetracycline resistance (TetR). Bacterial cultures are exposed to sub-MIC levels of antibiotics to induce controlled stress without significant cell death, followed by selection on tetracycline plates to isolate resistant revertants. The mutation frequency is quantified to assess the extent of stress-induced mutagenesis under sublethal antibiotic conditions. This work aims to elucidate the mechanistic link between antibiotic stress and the emergence of resistance, providing a foundation for strategies to limit mutagenesis and combat multidrug resistance.

RESEARCHERS: Jiya Arora, Emerald High School ‘28; Prisha Baruah, Piedmont High School ‘28

ADVISOR: Shee Lab, Molecular Biology, Genetics
KEYWORDS: Antibiotic resistance | Genetic Mutations | Public Health | Mutagenesis | Double Strand breaks

Department of Computer Science & Engineering

Copper Hardness-Strength Constraint Factor Grain Size Dependence Variability as Function of Work Hardening

Copper’s hardness-strength constraint factor, used to estimate strength non-destructively, has been experimentally determined to be around 3 with little grain size variance at 8% strain. It is important to understand how the constraint factor C changes at different levels of deformation. Current research shows that C may not be entirely constant but dependent on various strengthening and hardening mechanisms of materials, such as eutectic strengthening. C has also been shown to vary based on hardness testing methodology and the Poisson ratio. Previous research demonstrates that the relationship between C and grain size varies between yield and flow stresses. Presented here is an analysis of how the Hall-Petch relation changes as a continuous function of strain, used to then amend the definition of C to become a function of both grain size and strain. Current results indicate that σ0 changes in the form of a work hardening power law Aεn while Kσ decreases linearly with strain. As a 3-dimensional function, current work indicates that the sensitivity of C to grain size increases with decreasing strains, and changing strains can change the relationship of C to d from positive to negative. 

RESEARCHERS: Zachary Kwak, The King's Academy '27

ADVISOR:  Starostina Lab, Materials Science
KEYWORDS: Constraint factor | Work Hardening | Grain Size | Nondestructive Testing | Copper

Department of Chemistry, Biochemistry & Physics

Comparative study on three machine learning models in novel autonomous drone-based analysis of bike lane infrastructure  

Bike lanes help promote an eco-friendly form of transportation, necessary in a time when carbon emissions are high. Monitoring them to ensure minimal accident is necessary, but it can be time consuming. In this research, drones are utilized to effectively inspect bike lanes using a camera and sensor. Through the drone, we hope to efficiently find areas with more hazards in the infrastructure. By focusing on things like potholes and excess vegetation, less causalities will happen on the road, ensuring the safety of bikers. Using a drone is also cost efficient and functional, compared to manually inspecting bike lanes. The result of this project will lead to bike lanes being safer to ride on, promoting bicycling as it not only benefits the environment but also the health of a person. With the quadcopter drone, we will look for degradation on the bike lane using a camera and sensors. From here, the areas that were identified will be reported to road services.

RESEARCHERS: Samyuktha Bodduluri, McNeil High School '27

ADVISOR: McMahan Lab - Quantum Computing, Computer Engineering
KEYWORDS: Drone Technology | Bike Lane Safety | Infrastructure Monitoring | Hazard Detection | Sustainable Transportation

Department of Computer Science & Engineering

Machine Learning in Cuboidal and Spherical Nanoparticles: Merging ML Models for Improved Accuracy in SEM Image Analysis

Accurate nanoparticle identification and characterization from scanning electron microscopy (SEM) images remains challenging due to variations in particle morphology and background features. This work presents an integrated computational framework that combines optical character recognition (OCR), image processing, and YOLO-based object segmentation. A YOLO model is trained to identify and localize individual nanoparticles in SEM images, while OCR extracts scale bar information to convert pixel-based measurements into physical units. Image processing is subsequently applied to detected particle regions to characterize particle morphology and estimate physical areas. The proposed framework integrates multiple computational methods and multiple models into a unified workflow, enabling automated and reproducible quantitative characterization of nanoparticles from SEM images while reducing reliance on manual image analysis.

RESEARCHERS: Carter Tsao, Fairfax Senior High School '27

ADVISOR:  Starostina Lab, Materials Science
KEYWORDS: Scanning Electron Microscope | Machine Learning | YOLO | Nanoparticles | Computer Science

Department of Chemistry, Biochemistry & Physics

Antioxidant Activity of Fluorinated Edaravone Analogs for the Treatment of Amyotrophic Lateral Sclerosis  

Amyotrophic lateral sclerosis (ALS) is a rapidly progressive neurodegenerative disease characterized by the degeneration of upper and lower motor neurons. Although the precise molecular mechanisms underlying ALS remain incompletely understood, oxidative stress has been strongly implicated in disease progression. One therapeutic approach to mitigating oxidative damage involves the use of small-molecule antioxidants such as the FDA-approved drug Edaravone. Edaravone functions as a free radical scavenger that neutralizes reactive oxygen species (ROS), helping to reduce oxidative stress and mitochondrial damage associated with neuronal degeneration. The antioxidant activity of aromatic compounds is highly dependent on the electronic properties of ring substituents. To systematically examine these effects, we synthesized a focused library of edaravone analogs featuring electronically varied substituents. These derivatives include a compound in which the native methyl group is replaced by a strongly electron-withdrawing trifluoromethyl substituent, along with additional para-substituted phenyl variants. The in vitro potency of these analogs will be evaluated using co-dose DPPH, ABTS, and MTT assays to assess antioxidant activity and cell-line-specific viability. Using the resulting data, we aim to construct a Hammett correlation to quantify how the electronic properties of functional groups influence the radical-scavenging ability of edaravone analogs. 

RESEARCHERS: Riya Veera, Milpitas High School '28

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

Department of Biological, Human & Life Sciences

Evaluating the Influence of Hypoxic Stress on Escherichia coli Viability and Rifampicin Resistance

Understanding antibiotic resistance evolution requires examining how environmental conditions influence mutation and selection. Most studies focus on rapidly growing bacteria under nutrient-rich, well-oxygenated conditions, which may not reflect prolonged infection environments. Here, we used longitudinal, non-pooled sampling over 14 days to examine rifampicin resistance in Escherichia coli MG1655 under normoxic and hypoxic conditions. Rifampicin-resistant colony counts and mutation frequencies were monitored throughout prolonged incubation. In both conditions, absolute resistant colony counts declined over time, consistent with reduced population viability and potential fitness costs associated with rpoB mutations. However, mutation frequencies displayed contrasting patterns. Under normoxia, mutation frequencies progressively decreased, whereas hypoxic cultures showed increasing mutation frequencies over time. These findings suggest that oxygen availability can substantially alter the balance between bacterial survival, selection, and persistence of resistant variants during prolonged stress. Our results highlight the importance of hypoxic environments in resistance evolution and motivate further investigation of compensatory mutations, DNA repair pathways, and gene-expression changes under prolonged hypoxia.

RESEARCHERS: Lina Vernik, Homestead High School '28; Reya Kavuru, Homestead High School '28

ADVISOR:  Shee Lab, Molecular Biology, Genetics
KEYWORDS: Antibiotic Resistance | Rifampicin | Microbiology | RpoB Mutations | Transcription Inhibition

Department of Chemistry, Biochemistry & Physics

Installation of electrophilic alkyl bromides modulate the anticancer activity of palbociclib towards achieving isoform-selective CDK inhibition

Given their central role in aberrant cell cycle regulation found in many cancer types, cyclin dependent kinases (CDKs) have become an increasingly important target in the development of anticancer therapeutics. Recent efforts in this space have led to the discovery and development of palbociclib and ribociclib, two structurally homologous inhibitors of the closely related isoforms CDK 4 and 6. While these compounds have demonstrated remarkable clinical efficacy, potency, and safety tolerability profiles, their limited selectivity between CDK 4 and 6 posed unique target tolerability challenges. To address this challenge, and in efforts towards the development of similar compounds with improved CDK 4 and 6 isoform discrimination, several have elaborated a solvent exposed piperazine handle to either install a covalent warhead or other functional motifs that may imbue greater isoform selectivity. Inspired by these and other approaches, our laboratory has prepared a series of alkyl, alkyl halo, alkynyl amide, and sulfonate derivatives. Here we disclose the chemical synthesis of over a half dozen palbociclib analogs with both alkylating and non-alkylating motifs on their shared piperazine handle and current progress towards isoform selective biological applications of these compounds as therapeutic leads for the treatment of cancer.

RESEARCHERS: Jonathan Mazor-Hoofien, Palo Alto High School '27

ADVISOR: Njoo Lab, Synthesis | Physical Organic Chemistry | Catalysis | Chemical Biology | Spectroscopy | Medicinal Chemistry
KEYWORDS: Organic Synthesis, Medicinal Chemistry, CDK4/6, Kinase Inhibition, Covalent Inhibition