🟢 Open to full-time roles · AI/ML Engineering · Backend · LLM Infra · Batch 2025
$ whoami
Shubham Singh · bihari-bhau · Gurugram, India 🇮🇳
$ cat current.txt
LLM Post-Training Intern @ Ethara AI
↳ Built Kaiju — AI coding agent benchmark pipeline (Commit0 / ICLR 2025)
↳ Evaluating GPT-4o, Claude 4.7, Gemini across 100+ Python repos
↳ 500+ RLHF samples annotated · 38 eval criteria · 6+ LLMs benchmarked
↳ Journey: EEE → Full-Stack → AI/ML Engineering 🚀Benchmarks AI agents on reconstructing Python libraries from scratch. Built on Commit0 (arXiv:2412.01769 · ICLR 2025)
GitHub Repos → AST Stripper → Stubs → AI Agent → pytest → Score
(2000+ ⭐ (function bodies (empty (Claude / (pass (ethara
80%+ Python) stripped → ∅) shells) GPT-4o / …) rate) splits)
| Split | Libraries | Purpose |
|---|---|---|
ethara |
8 libraries | Full benchmark |
ethara-lite |
4 libraries | Lightweight eval |
Languages
Backend & Frontend
AI / ML
DevOps & Tools
| Project | Stack | What it does | Live |
|---|---|---|---|
| 🦖 Kaiju | Python · AST · pytest | AI coding agent benchmarking pipeline (Commit0 / ICLR 2025) | — |
| 📊 rlhf-eval | React · FastAPI · PostgreSQL · Docker | Full-stack RLHF dataset builder — pairwise comparisons, JSONL export | — |
| 🧰 LLM Toolkit | Next.js · TypeScript · Tailwind · Supabase | Modular toolkit for LLM prompt experiments, evals, and dataset workflows | 🔗 |
| 🎯 Lead Sniper | n8n · LLM · Slack/Discord | GitHub stargazer → enrichment → LLM pitch → auto-delivery | — |
| 🌦 Weather-Aware Order Checker | Node.js · OpenWeatherMap API | Order decisions driven by real-time weather via Promise.all |
— |
| 📚 Bihar Skill Hub | React · HTML · CSS | Ed-tech platform bridging Bihar's skill gap | 🔗 |
| 🍽 Meal-Buddy | Python · Django · FastAPI | Meal planning and suggestion API | — |
┌────────────────────────────────────────────────────────────────────┐
│ 🤖 LLMs Evaluated → 6+ (GPT-4o, Claude, Gemini…) │
│ 📦 Python Repos Processed → 100+ (repo_finder.py pipeline) │
│ ✅ Eval Criteria / Repo → 38 (fully automated) │
│ 📝 RLHF Samples Annotated → 500+ │
│ 🏗 Custom Benchmark Splits → 2 (ethara / ethara-lite) │
└────────────────────────────────────────────────────────────────────┘
education = {
"degree": "B.Tech — Electrical & Electronics Engineering",
"college": "Sershah Engineering College, Bihar",
"batch": "2025",
"training": "Java Full Stack @ JSpiders, Noida",
"journey": "EEE → Full-Stack Dev → AI/ML Engineering 🚀",
}
"Ship it. Benchmark it. Improve it."