Featured Projects
A collection of things I've built
12 projects across AI, Healthcare & Data Engineering
Multi-Agent Clinical Decision Support System
A multi-agent AI clinical decision support system for analyzing patient data, retrieving medical context, and generating structured care insights.
ReasonFlow
Multi-agent orchestration system built on LangGraph with router-driven execution, dependency-aware parallel specialist dispatch, reflection loops, and optional human-in-the-loop approvals.
AI Business Assistant
Multi-step RAG pipeline with LangGraph for PDF Q&A. Supports voice input via Whisper and text chat with transparency into retrieved chunks.
AWS ETL Data Pipeline
End-to-end pipeline to manage, streamline, and analyze YouTube trending data based on categories and metrics.
SQL Data Warehouse
Comprehensive data warehousing solution demonstrating industry best practices from schema design to actionable analytics.
RelGraph
RelGraph trains on multi-table relational data via RelBench: rows and foreign keys become a temporal graph, with predictions at each entity and timestamp and no future leakage. It benchmarks a flat sklearn baseline, GraphSAGE, and RelGT-lite, a from-scratch Relational Graph Transformer checked against the official implementation. The repo also includes a PQL-style label compiler (pandas-validated) and link prediction. Entity-task numbers are validation scores on a CPU subset, not RelBench leaderboard test results.
Antibiogram Zone Detector
Automated analysis of bacterial susceptibility zones using image classification, object detection, and pixel-level segmentation.
Research Based Thyroid Nodule Classification with Explainable AI
End-to-end medical imaging review project that accepts a thyroid ultrasound still image, runs TensorFlow-based binary classification, generates a Grad-CAM attention map for interpretability, and exports the result as a DOCX report.
LangChain EDA Assistant
Chatbot for natural language interactions with SQL databases, enabling real-time visualization and data exploration insights.
Multimodal Food Recommendation RAG
RAG system combining text and image understanding for personalized recipe suggestions with explainable AI.
ImageTalk
Multimodal AI app for voice-based interaction over visual content using Groq, Whisper, and LLaMA for end-to-end speech understanding.
Pneumonia Detection
Deep learning pipeline using modified ResNet18 for chest X-ray classification with Class Activation Maps for visual interpretability.