Production voice-first triage platform that pre-authorises emergency healthcare payments into escrow before the patient arrives. Built with Next.js 16, TypeScript, Interswitch payment rails (OAuth 2.0, HMAC-SHA-512 verified webhooks), and three real-time SSE dashboards coordinated over PostgreSQL. Built for the Enyata × Interswitch Buildathon.

Distributed microservices backend built from scratch: face recognition engine, liveness detection, exam session manager, real-time event bus — all communicating via gRPC with fault-tolerant inter-service contracts. Full Terraform IaC for GCP, pgvector for face embedding similarity search at scale.

Autonomous research agent platform with Golang (Modus) backend processing intensive data pipelines, connected to a Next.js frontend via GraphQL. Multi-service architecture with real-time results streaming.

Enterprise LLM platform drafting and executing legal agreements with DocuSign API integration. Fault-tolerant API design for high-stakes document workflows at production scale.

Decentralized health data platform on blockchain. Architected and deployed end-to-end in a 48-hour hackathon sprint — full distributed architecture with health records tracked via DAG-based consensus.

Privacy-preserving blockchain voting system built for the NDPC Code4Privacy Hackathon. Solidity smart contracts with zero-knowledge voting primitives. Winner of the hackathon.

Building an AI-driven remote cancer diagnostic platform using smart cameras to transmit high-fidelity pathology images, enabling expert oncologist review from resource-limited settings.

Trained and deployed volumetric segmentation models (U-Net) for automated tumor boundary delineation from CT scans. Built preprocessing pipelines for DICOM parsing and 3D patch extraction.

Built machine learning pipeline using Google's Pathology Foundation Model for H&E-stained microscopy analysis. Achieved 89% accuracy distinguishing malignant from benign breast tissue.

Authored and released a PyTorch-native open source library implementing metaheuristic optimization algorithms. Published as a preprint on arXiv after passing peer review.

Engineered an agentic voice operating system specifically tailored for healthcare applications. Designed to handle automated patient triage, hospital resource management, and clinical workflows.

Developed a comprehensive web and mobile application that ingests standard YouTube videos and dynamically transforms them into structured, interactive educational courses.

Architecting a novel EdTech platform that leverages Large Language Models to teach complex concepts by generating direct, personalized analogies based on fields the user already understands.

Medical AI that uses Cohere classifier to classify the intent of social media posts, understand emotional context, and generate contextual message responses. Fine-tuned Cohere classification model on social media tweet dataset.

Real robot that sorts waste using computer vision. 3D printed open-source model, trained classification model for 4 categories (metal, glass, paper, plastic). Arduino and Raspberry Pi for movement and sorting logic.

AI buddy for chatting with your data as a business and getting real insights. Built for Google AI Hackathon 2024 using Gemini models for natural language data querying and analysis.

AI-based platform for analyzing influence of stock news on the stock market and making financial decisions. Streamlit frontend, LangChain for the agent, Gemini for LLM, MongoDB for vector storage and retrieval, Finnhub API for market data.

Agentic application backend with voice-first interface for restaurant and hotel recommendations. Implemented user preference tracking using Neo4j knowledge graph database. Built RESTful APIs with FastAPI ensuring low-latency responses.

Blockchain platform on Solana enabling game developers to create decentralized game assets. Developed backend API with NestJS handling asset creation and blockchain transaction management. Created Python CLI tool using Typer for seamless integration.

Full-stack Django application converting YouTube videos into blog articles using AI. Integrated YouTube Transcript API and AssemblyAI for audio-to-text transcription with 95%+ accuracy. Implemented Google Generative AI for intelligent content generation.