Shipping production AI products end-to-end with FastAPI, TypeScript, PostgreSQL, Redis, and LLMs.

github

Aditya Mondal  • Peerlist

Built a full-stack arbitrage screener that detects price gaps between NSE and BSE for Indian large-cap stocks in real time. The backend runs two async services — a price fetcher (yfinance + ThreadPoolExecutor) and an arbitrage engine — decoupled via Redis, with live prices streamed to the browser over WebSocket. Deployed on Render + Vercel with Supabase (Postgres) and Upstash (Redis).

ArbScreen — NSE/BSE Real-Time Arbitrage Scree

A parallel graph processing engine leveraging CUDA for GPU-accelerated traversal and computation. Designed to benchmark performance characteristics of graph algorithms under parallel execution vs. sequential CPU baselines. Engineering focus: CUDA kernel design, memory coalescing, parallel BFS/SSSP, performance profiling and benchmarking. Relevant for: high-performance computing, systems engineering, infra-heavy and compute-intensive workloads.

GPU-Accelerated Graph Engine

Developer tooling that uses an LLM to analyze CI test failure logs and surface root cause hypotheses. Designed to reduce debugging time in fast-moving engineering teams. Engineering focus: log parsing pipeline, LLM prompt engineering for structured diagnostic output, integration with CI artifact formats. Relevant for: developer tooling, AI infra, observability, engineering productivity startups.

LLM-Powered Test Failure Analyzer

RAG-powered macro market analyst chatbot. Semantic search over financial data using Supabase pgvector and Transformers.js local embeddings. Streaming responses via Claude API SSE. Built in Next.js 15 + TypeScript, deployed on Vercel. Engineering focus: low-latency token rendering via RAF batching, in-browser embedding inference (no external embedding API call), vector similarity retrieval.

MacroLens AI

Real-time price divergence detector across NSE and BSE. FastAPI async backend with WebSocket-based live price feeds. React frontend with live table updates. Fully containerized with Docker. Engineering focus: async polling architecture, cross-exchange normalization, real-time push to client.

NSE/BSE Stock Arbitrage Screener

Document intelligence pipeline for extracting and structuring heading hierarchies from arbitrary PDFs. Useful for automated document indexing and knowledge base construction. Engineering focus: layout-aware parsing, heuristic classification of heading levels, structured output generation.