Pushkal Pratap Singh

I’m an AI/ML and backend engineer who builds LLM gateways, RAG pipelines, and ML microservices that work reliably in production. I focus on practical performance—latency, cost, and observability—while keeping model behavior aligned to real task needs.

In my backend work, I developed and maintained REST API endpoints and database schemas in production, collaborating closely with frontend engineers for feature delivery, bug triage, and code review across sprint cycles. I also evaluated LLM outputs across coding, reasoning, and math, writing prompt-response pairs and rubric-based assessments to improve alignment.

My standout projects include Nyay Samiti, where I fine-tuned multiple transformer models for Indian legal intelligence and shipped a FastAPI ML microservice with a sequential inference pipeline. I also built a production-grade Prompt Intelligence Layer (PIL) LLM gateway with intent-aware routing, semantic caching, PII masking, provider fallbacks, and trace-level instrumentation.

I enjoy turning prototypes into dependable systems—like ZIT, a serverless AWS Lambda backend with a RAG pipeline using FAISS and DynamoDB metadata. Beyond engineering, I was a Smart India Hackathon 2025 National Finalist and an AI For Bharat 2026 (powered by AWS) 2nd Runner-Up.