I build backend systems that think.
Currently engineering production GenAI pipelines at Pocket FM, where my team curated 500K+ hours of audio and text data for model training. Previously: backend engineering at ADP (Spring Boot, Kafka) and applied DL research at LNMIIT.
Graduating June 2026. Open to full-time SDE, AI Engineer, and GenAI roles at product companies and AI startups, plus collaborations on AI memory, low-resource NLP, audio ML, and dev tooling.
🌐 subarnasaikia.vercel.app · ✉️ LinkedIn · 🐦 X · 📊 Kaggle
| Project | What it is | Stack | Status |
|---|---|---|---|
| Genesis | Open-source NLP annotation platform for coreference resolution. Used by university research groups working on Assamese low-resource language models. | Java · Spring Boot · React · OAuth2 · Docker | 🟢 Active |
| Loam | A journal that pulls you back daily through intrinsic wonder, not guilt. | Rust | 🟢 Active |
| kaggle-skills | A toolkit of Claude Code skills, hooks, and scripts that compound your Kaggle knowledge across competitions. | Shell · Python · Claude Code | 🟢 Active |
| NoteX | Context-aware AI note-taking — semantic search, auto-revision, dynamic quizzes. | Next.js · LangChain · OpenAI · MongoDB | 🟡 Maintained |
| Solar Efficiency Prediction | Top 25 globally (out of 1,500+ teams) — Zelestra × AWS ML Ascend Challenge. | Python · CatBoost · Ensemble Learning | ✅ Shipped |
| ChessEngine | Chess engine from scratch — bitboard representation, SDL2 GUI, evolving AI bot. | C++ · SDL2 · Bitboards | ✅ Shipped |
| Notebook | Competition | Topic | Upvotes |
|---|---|---|---|
| Reversal Points in US Equities | Detecting Reversal Points (rank 59 / 426) | Time-series · stock data | 🔥 74 |
| CAFA 6 — base model | CAFA 6 Protein Function Prediction | Bioinformatics · multi-label | 59 |
| Multi-Label Dense NN — Transformer Model | CAFA 6 | Transformer · multi-label | 34 |
| It's Backpack time — XGBoost | Backpack Prediction (Playground S5E2) | Tabular · gradient boosting | 20 |
| S5E11 — EDA + LightGBM | Predicting Loan Payback | Tabular · EDA | 12 |
| Competition | Result | Field size |
|---|---|---|
| Detecting Reversal Points in US Equities | 59 / 426 🥉 | Community |
| Backpack Prediction Challenge (S5E2) | 507 / 3,393 | Playground |
| Diabetes Prediction Challenge (S5E12) | 395 / 4,206 | Playground |
| Predicting Loan Payback (S5E11) | 632 / 3,724 | Playground |
| FIDE & Google Efficient Chess AI | 838 / 1,120 | Featured |
| Predicting Irrigation Need (S6E4) | 1,389 / 4,315 | Playground |
- 🥈 Codeforces Specialist (max rating 1432) · @Subarna1
- 🌟 CodeChef 3⭐ (max rating 1670) · @subarna1
- 🏅 Kaggle Notebooks Expert · best rank 921 / 61,556 · 8 medals · @subarnasaikia
- 🎯 GATE CS '26 qualified
- 📜 Oracle ACE Apprentice · OCI 2025 AI Foundations Associate certified
Personal AI memory · Audio ML · Low-resource NLP · Dev tools that compound · Production LLM systems · Anything where the bar is "actually works at scale"
📬 Open to roles, collabs, and good conversations.