Portfolio – Omer Chaudhry

Works

My Projects

Smart IMS GenAI System

Developed an AI-powered warehouse management platform integrating multi-agent orchestration (SQLite, AutoGen AI) to
automate inventory tracking, order fulfillment, and workflow execution.

Fine-Tuned DistilBERT for Question Answering (QA) System

Developed an extractive QA model using fine-tuned DistilBERT, implementing custom tokenization & span ranking to
achieve ~65% F1-score on the Natural Questions dataset

AttnSleep: Attention-Based CNN Intensive Sleep Staging Model

Optimized AttnSleep deep learning model for EEG-based sleep stage classification (TensorFlow, CNN, Multi-Head
Attention), achieving 80% categorical accuracy & 70% F1-score.

Passport Index & Economic Indicators Analysis

Conducted ML-driven economic analysis integrating logistic regression & clustering to assess the correlation between
passport strength, GDP, literacy rate, and education levels.

RAG based Chatbot with OpenAI/DeepSeek Integration

Developed a RAG-based customer support chatbot for an e-commerce site using OpenAI and DeepSeek LLMs, enabling intelligent product search and troubleshooting for appliance parts. Built a modern React frontend and scalable backend with vector database integration for accurate, extensible query resolution.

scGrapHiC with GAT, Transformer Encoder and CNNs

Contributed to scGrapHiC, a deep learning framework integrating Graph Attention Networks, transformer encoders, and positional encodings to predict pseudo-bulk single-cell Hi-C (scHi-C) contact maps from scRNA-seq data increasing accuracy by 15%, across 1M+ cell samples.

For more projects related to GenAI, Machine Learning, Data Science, and Full Stack development click PORTFOLIO