Hitaansh Jain

Software Engineer Intern

• Shipped full-stack customization for AI-generated briefings (per-section reorder, toggles, and detail control) to production, refactoring generation from a hardcoded prompt to dynamically assembled sections backed by a JSONB section model.
• Built Clear Tabs (bulk-closes a user's open tabs and returns a briefing summarizing them, shipped as Link Bankruptcy) formatting end-to-end: built a paginated, authenticated FastAPI history endpoint with a React Native screen, a submit-time format picker with user-written sections, and preserved submission metadata a scheduled cleanup job would otherwise have deleted.
• Built a multi-phase backend lifecycle system across five stacked PRs, covering schema design, a state machine, the write API, email delivery, and agent-facing parity, plus the web UI in the final PR.
• Wrote 700+ automated tests on a test-per-feature workflow across a multi-provider LLM pipeline, FastAPI/PostgreSQL backend, and React Native/Expo frontend, with every change merged through a senior engineer's code review.

Software Engineer Intern, AI (Part-Time)

• Built a five-stage TypeScript/Next.js pipeline that turns a calculus textbook problem into three verified study artifacts: a compiled LaTeX case study, concept flashcards, and a step-by-step solution walkthrough, deployed for internal review with the content handed to MathGPT's engineering team for integration.
• Designed a generator/critic LLM architecture where the critic re-solves each problem before the generator's drafts even exist, and halts the pipeline on mismatch.
• Owned the corpus extraction system, stage validation contracts, and MySQL flashcard cache, covered by 1,000+ automated tests including negative controls, working on a three-intern Agile team with weekly project-lead syncs.

Software Engineer Intern (Part-Time)

• Backend CRUD and a full MySQL → Microsoft Access migration for an inventory system serving 1,500+ retail locations, moved to Access to meet the client's offline, zero-infrastructure deployment constraint (394-column schema, 700+ queries rewritten across 43 files).