RAGenetics – Privacy-Preserving RAG for Genetic Test Reports (Link)

Tools: Python, Differential Privacy, RAG, LLMs, VectorDB 

• Built a privacy-preserving Retrieval-Augmented Generation (RAG) system for genetic test report queries.

• Implemented two Differential-Privacy algorithms:

1. DPVoteRAG: sample-and-aggregate with Report-Noisy-Max token release.

2. DPSparseVoteRAG: Sparse Vector Technique-gated RAG to reduce privacy cost when voters align.

• Supports synthetic data generation for safely testing genomic-text pipelines.

• Designed modular configs, vectorstore builder, voting inference & privacy budgeting.

• Ensures DP-secure token generation — research-grade evaluation only (non-clinical use).

PneumoGAN – Pneumonia Detection using GAN Variant (Link)

Tools: Python, GAN, Deep Learning, Medical Imaging 

• Generated synthetic X-rays using a custom GAN architecture.

• Achieved 94.99% accuracy in pneumonia detection.

• Used 5,000+ images with augmentation for robust training.

• Evaluated using precision, recall, and F1-score for medical reliability.

CashShield – Fake Currency Detection using ML (Link)

Tools: MATLAB, Image Processing, ML Classification

• Built a system to detect counterfeit Indian currency using ML-based feature extraction.

• Achieved 95%+ accuracy in detecting fake vs original notes.

• Classified currency denominations with 90%+ efficiency using ROI intensity analysis.

Telecom Complaints Monitoring System

Tools: ARIMA, LSTM, XGBoost, ETS, Python  

• Forecasted complaint trends for telecom sectors using time-series ML.

• ARIMA outperformed other models during comparative testing.

Housing Price Prediction (Link)

Tools: Python, Pandas, NumPy, scikit-learn, Seaborn, Matplotlib  

• Built a machine learning model to predict real-estate prices using structured housing data.

• Performed data preprocessing, handling missing values, outliers and categorical encoding.

• Conducted Exploratory Data Analysis (EDA) to study feature relationships & correlations.

• Implemented regression-based models for price prediction with performance evaluation using RMSE & R².

• Created visual trends & distribution plots for actionable insight into pricing behavior.

Cricket Analytics – Batting Performance Analysis & Prediction (Link)

Tools: R, ggplot2, dplyr, Statistical Modelling, EDA   

• Analyzed international batsmen performance using historical ICC cricket datasets.

• Performed data cleaning, preprocessing & feature engineering on batting metrics.

• Conducted Exploratory Data Analysis (EDA) on averages, strike rate, 100s, 50s & dismissals.

• Built Multiple Linear Regression models to predict batting average trends.

• Visualized global player distribution, run patterns & year-wise performance trends.

Retail Sales Analytics (Link)

Tools: PostgreSQL, SQL, EDA, Data Visualization  

• Built & cleaned a retail sales DB with 2000+ records.

• Queried top-spending customers, peak sales months & buying patterns.

• Delivered actionable insights to improve store revenue strategy.

Sorting Memory Database (Link)

Tools: Java, Julia, Linked List DB, Performance Benchmarking 

• Built an in-memory toy database using a linked-list storage structure for student records.

• Implemented two sorting algorithms — Bubble Sort & Insertion Sort — for performance comparison.

• Designed CLI workflow to load, sort, export, and continue operations recursively.

• Compared CPU time, memory usage, and disk I/O using Ubuntu profiling tools (/usr/bin/time, iostat).

• Results showed Insertion Sort outperforming Bubble Sort consistently in memory-bound execution.

Power-Efficient 4–2 Compressor for Tree Multiplier (Link)

Tools: Cadence Virtuoso, Low-Power VLSI Design 

• Designed a 4-2 compressor optimized for lower power consumption.

• Achieved 35.18% lower power vs existing architectures.

• Implemented using custom XOR gates with AVLS, LECTOR & Adiabatic techniques.