$ whoami

Jay Gupta

I build agent systems and the infrastructure they run on.

AI engineer @ encraft · building sketchpen (early access)

// about

Final-year Computer Science (Data Science) undergrad who has spent the last two years shipping AI-heavy products in production: async backends, ETL pipelines, agents, retrieval systems, and the unglamorous parts underneath them.

I care about systems that hold up when real users show up. Correctness first, cost always on the table, tradeoffs made deliberately.

// now

encraftvideo-analysis & retrieval pipelines for enterprise compliance
sketchpenwhiteboard-video engine, built from scratch (early access)
ossagent tooling: openchamber, ai-python

// experience

AI Engineer @ Encraft

2026 – now

Building video-analysis and evaluation pipelines for enterprise compliance: resumable stage machines, structured outputs, retry-safe orchestration.
Hybrid RAG over regulatory documents, and rubric-based evaluation harnesses for model output.

AI Engineer @ Prepairo (Vertune Data Tech)

Dec 2024 – Jan 2026

Owned a daily end-to-end ETL system (scrape → normalize → summarize → store) that contributed to a 3× increase in app downloads and 60% DAU growth.
Shipped a production GRE AI tutor on pydantic-ai: schema-validated outputs, tool-based reasoning, persistent user context.
Built the async backend for the GRE modules (FastAPI, async SQLAlchemy, PostgreSQL) with predictable latency and clean domain separation.
Engineered subscription infrastructure with Razorpay: verified webhooks, idempotent flows, real-time entitlements.
Cut AI infrastructure cost ~50% via batch inference, prompt restructuring, and model selection by cost–quality.

// work

Sketchpen · AI whiteboard-video engine

Script in, animated explainer out. Voiceover with word-level sync and a hand-drawn rendering engine built from scratch. Multi-language, near-real-time. Twenty-one iterations in seven working days. Live in early access.

next.js · fastapi · pydantic-ai · custom python renderer

Apex · desktop AI workspace

A Cursor-like desktop assistant: local notes, document context, type-safe agents over MCP, Rust filesystem service. Built because I wanted an AI environment I would actually use. Private for now.

tauri v2 · react · rust · pydantic-ai · chromadb · mcp

// open source

openchamber #2796 · /btw: side questions in a temporary forked session
openchamber #2767 · nested git repositories across the git surfaces
ai-python #260 + #259 · media-type detection and stream-finish fixes

// stack

languages: python · typescript · rust (basics) · sql
ai: agents · rag · evals · tts · ocr · voice
backend: fastapi · postgres · async sqlalchemy · docker · aws · gcp
product: next.js · react · tauri · tailwind

// education

B.Tech, Computer Science & Engineering (Data Science), RCOEM Nagpur · 2023–present · CGPA 8.08