Skip to content
View kansalshivam's full-sized avatar

Highlights

  • Pro

Block or report kansalshivam

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
kansalshivam/README.md

Shivam Kansal

Backend and AI systems engineer building the layer between raw databases and people who don't want to write SQL to talk to them.

Currently interning at CRIS (Centre for Railway Information Systems), where I'm putting a RAG middleware layer in front of a live railway operations database — the kind of system where "just hardcode the query mappings" stops being an option after about the third schema. Outside of work: Top 800 of 32,000+ teams, solo, at Meta × Scaler × HuggingFace's OpenEnv hackathon, reviewed by Meta's engineering team at the 48-hour Bangalore finale.

Currently working on: schema variability handling for the RAG planning layer at CRIS — separating "the model decides what to ask" from "the model touches the database," which turns out to matter a lot more than it sounds.

Graduating 2027, open to SDE / AI engineering internships now. If you're building something involving LLMs, RAG, or backend systems that need to hold up under real load, I'd like to hear about it.

Work that's shipped

AI Engineer Assessment Platformlive A hiring platform that scores candidates on how they debug a broken RAG pipeline, not whether they can center a div. Multi-tenant, JWT-secured, four distinct simulated failure modes, full session replay via a telemetry engine with monotonic timestamps. Deployed on Vercel + Render with Docker multi-stage builds and GitHub Actions CI/CD. Built solo. React TypeScript FastAPI PostgreSQL Docker JWT

Schema Migration GymOpenEnv AI Hackathon, Top 800 / 32,000+ A Gymnasium-compatible RL benchmark for SQL schema migration. Six task types, three difficulty tiers, and adversarial cases specifically designed to break greedy agents (zero-overlap renames, constraint repair traps). Each submission gets its own fresh Postgres container via the Docker SDK, diffed across five dimensions. The heuristic baseline scores a perfect 1.0 on standard tasks and drops to 0.48 on the adversarial set — showing exactly where rule-based logic stops working and reasoning has to take over. Python FastAPI PostgreSQL Docker OpenEnv Gymnasium

Pollution Response DashboardDelhi Govt Hackathon, Top 6 Finalist A planning dashboard for municipal pollution-response deployment, built from idea to finalist pitch. React TypeScript

FOSSEE Workshop Bookingopen-source, IIT Bombay Mobile-first UI/UX overhaul of the FOSSEE workshop booking platform, built to degrade gracefully across devices. React Django

Skills

Languages Python, TypeScript, JavaScript, SQL
Backend FastAPI, Django, REST/OpenAPI 3.0, LangChain, RAG pipeline design
AI/ML LLM API integration, prompt engineering, HuggingFace Transformers, OpenEnv, Gymnasium
Data PostgreSQL, SQLite, schema design & migration
Infra Docker, GitHub Actions, Vercel, Render

Also shipped applied ML work — a dynamic pricing model — through IIT Guwahati's Summer Analytics 2025. Currently deep in DSA fundamentals (trees, graphs, DP) alongside the systems work above.


Get in touch

LinkedIn · Email · X

Pinned Loading

  1. AssessAi AssessAi Public

    Python 2

  2. fossee-workshop-booking fossee-workshop-booking Public

    UI/UX enhancement of FOSSEE Workshop Booking platform using React + Django (mobile-first, progressive enhancement)

    Python 2

  3. pollution-navigation-system pollution-navigation-system Public

    A React and TypeScript planning dashboard for municipal pollution-response deployment.

    TypeScript 2

  4. schema-migration-gym schema-migration-gym Public

    Python 2