I’m Prathamesh Prasad, a developer based in West Bengal, India. My focus is backend engineering and applied AI, from designing reliable software services to building and evaluating AI-powered applications.
A research assistant that answers questions over documents with citations. The Express backend handles authentication, rate limiting, and pgvector retrieval. It uses PostgreSQL and Redis for application state and streams progress to the client.
A FastAPI service runs the LangGraph workflow. It rewrites weak queries, checks answers against retrieved evidence, and abstains when the evidence is insufficient.
I benchmarked dense retrieval against BM25 and cross-encoder reranking. Dense retrieval performed best on the project’s evaluation set and became the default. Live demo
Express PostgreSQL Redis pgvector Python FastAPI LangGraph MCP
A payment reliability testing tool built for the PayPal AI Hackathon. It reproduces checkout failures in PayPal Sandbox, reconciles payment and shop records, and reruns scenarios to verify fixes. Repository
A prompt-to-website app with a Node.js backend for streaming previews, version history, and Stripe-backed credit billing. Users can revise a site through chat, edit individual sections, and roll back changes. Repository
An event-driven benchmarking platform that compiles and stress-tests financial order-matching engines and streams their performance metrics. Repository
I contribute bug fixes and regression tests that improve document indexing and command-line reliability in XERJ, a Rust search engine for AI applications. XERJ repository
APIs and services
Data and infrastructure
Workflows and retrieval
Grounding, evaluation, and observability