Developed LLM-powered applications using Retrieval-Augmented Generation (RAG) pipelines, improving answer accuracy and reducing hallucinated citations through grounded web data.
Built and scaled ML pipelines and evaluation frameworks for model training, inference, and performance monitoring.
🏖️ Side Quests
Check out all the cool projects I've worked on!
qtypi - Python Library for Quantum Computing Operations
Developed a Python library for quantum state and gate manipulations, including common and custom gates
Developed a vision-language model to improve object count accuracy in text-to-image generation
Integrated Stable Diffusion with YOLOv8x for precise object detection and counting, and Low-Rank Adaptation (LoRA) for efficient fine-tuning in resource-constrained environments
Detecting and Mitigating Bias in Facial Recognition
Conducted comprehensive bias analysis in facial recognition models, primarily focusing on DeepFace and Viola Jones algorithms
Implemented a methodology to enhance DeepFace's comprehension of diverse female facial traits aiming to refine gender-specific facial interpretations and reduce observed biases
DeepRacer
Finalist for the AWS DeepRacer Women's League 2021
Developed a Reinforcement Learning based self-driving car racing agent using AWS DeepRacer
Adapted and iterated on Reinforcement Learning algorithms to optimize self-driving car racing performance, resulting in notable progress throughout the competition
Engineered a custom blockchain and cryptocurrency from the ground up, implementing innovative consensus mechanism for decentralized security
Architected and built out core cryptocurrency features including transaction processing, validation, and propagation across peer nodes
Established secure blockchain protocols enabling transparent and tamper-proof transaction ledgers
Sentiment Analysis on Text Feedback
Implemented advanced data preprocessing techniques, including text cleaning, tokenization, and feature engineering, to ensure optimal model performance
Designed a sentiment analysis system using Long Short Term Memory (LSTM) based Deep Neural Network to automatically analyze and classify user-generated text feedback
Engineered a unique gaming experience by developing a gesture-controlled version of the classic Snake game
Utilized Python and Computer Vision techniques to develop a gesture recognition system, enabling intuitive control of the game character's movements through real-time analysis of player's gestures