Pre-Final Year • B.Tech Data Science & AI • IIT Bhilai
I build systems from scratch — network protocols, distributed sync engines, database internals, and infrastructure that makes other engineers more productive. My philosophy: if you want to understand something, build it yourself.
Currently:
- OpenLake Coordinator managing 30+ repos and 100+ contributors
- MOSIP Contributor (selected thrice from 4,200+ applicants) building India's digital identity infrastructure
| Project | What It Is | Stack |
|---|---|---|
| BitTorrent Client | Full BEP-0003 protocol from scratch — bencode parsing, wire protocol handshakes, DHT via Kademlia, rarest-first piece selection, choking/unchoking with tit-for-tat, endgame mode | Go |
| ACORDE | Local-first P2P data sync with CRDTs (G-Counters, LWW-Element Sets), XChaCha20-Poly1305 encryption, Kademlia DHT, SQLite WAL mode | Go, libp2p, SQLite |
| B+ Tree | Production-grade from scratch — node splitting, sibling borrowing, doubly-linked leaf list for O(n) range scans, configurable order | Go |
| Kudos | Token-economy social platform — MLFQ-inspired feed ranking (from OS scheduling), progressive taxation, atomic transactions with SELECT FOR UPDATE, gRPC microservices |
Go, Node.js, PostgreSQL, gRPC |
| HalfLife | Temporal reranking for RAG — intent classification, multi-strategy decay (exponential/piecewise/learned MLP), +191% improvement on fresh-intent queries | Python, NumPy, Redis, Qdrant |
| Relix | Distributed data processing with checkpoint-based recovery, custom scheduling across heterogeneous workers, graceful degradation | Go, PostgreSQL |
| Project | What It Is | Stack |
|---|---|---|
| GitIntel | LLM-powered GitHub intelligence — transforms messy discussion threads into structured engineering digests with D3.js decision dependency graphs | Python, FastAPI, GraphQL, OpenAI, D3.js |
| RAGfolio | RAG system with semantic boundary chunking, recursive character splitting, intelligent passage ranking, citation tracking | Python, Vector DBs, HNSW |
| MetaIndex | ML-adaptive database index selection — meta-learning to predict optimal index configurations, SIMD/mmap optimization | Python, ML |
| Project | What It Is | Stack |
|---|---|---|
| AcadMap | Crowd-sourced course resource platform for IIT Bhilai — used by the student community | Next.js 15, TypeScript, Supabase |
Click to expand full stack
Languages: Go · Python · Rust · TypeScript · C/C++ · Java · JavaScript · SQL · Dart
Backend: Next.js · FastAPI · Node.js · NestJS · Spring Boot · Express · gRPC
Databases: PostgreSQL · MongoDB · SQLite · Supabase · Neo4j · Redis · Vector DBs · Firebase
Systems & Protocols: libp2p · CRDTs · Kademlia DHT · BitTorrent BEP · CBOR · COSE · OpenID4VCI · DID/VCs · GraphQL
DevOps: Docker · GitHub Actions · CI/CD · Linux · Hyprland · Tmux · Kubernetes
AI/ML: LLMs · RAG · PyTorch · Scikit-learn · Vector Embeddings · Meta-Learning · GNNs · Knowledge Graphs
Cryptography: XChaCha20-Poly1305 · SHA-256 · BitString Status Lists · W3C Verifiable Credentials · COSE_Sign1
Standards: ISO/IEC 18013-5 · W3C VC Data Model 2.0 · OpenID4VCI
|
🏛️ MOSIP Foundation (C4GT) Contributor · Apr 2025 – Present • Double selection: Sprint 2 + DMP 2025 (from 4,200+ applicants) • W3C-compliant credential revocation with BitString Status Lists • ISO/IEC 18013-5 mDoc implementation — CBOR, COSE, MSO, OpenID4VCI • Pre-Auth Code Flow (PR #483: 90 commits, 2,102 lines) • Total earned: ₹1,40,000+ |
🎓 FOSSEE, IIT Bombay Summer Fellow · May 2025 – Jul 2025 • OSDAG — Open Source Design and Analysis of Steel Structures • Contributed to democratizing engineering design tools • Large-scale collaborative development experience |
|
🌊 OpenLake, IIT Bhilai Coordinator · Apr 2024 – Present • Managing 60+ repositories with standardized CI/CD, labels, PR templates • 100+ contributors, 20+ juniors mentored, 10+ events organized • Designed 12-month matrix structure: 4 domain clubs × 4 verticals • Deep code audit of 19 repos via AST parsing — found 20+ critical issues • Drafted organizational constitution, authored Annual Report 2025-26 |
|
- Distributed Systems — ACORDE refactoring, consensus protocols, high availability patterns
- Database Internals — CMU 15-445 (Bustub), B+ Trees, learned indexes, query optimization
- OpenSSF Mentorship 2026 — RSTUF Visualizer, SBOMit + Trivy, gittuf TUI
- Rust — 5-phase learning path: content fetcher → tar clone → WhatsApp analyzer → CLI tool
- Build-Your-Own-X — BitTorrent ✓ · B+ Tree ✓ · Git (in progress) · Database · Search Engine · Redis · Shell
I don't just learn — I measure and analyze everything:
- 5,180 ChatGPT conversations analyzed over 3+ years (17,704 user messages)
- Custom Python pipeline clustering conversations into 12 cognitive modes (TF-IDF + KMeans)
- Behavioral analysis revealed: 1.95 debug-to-build ratio, 0.4% review rate, optimal learning at 9-10 PM
- 345,000+ WhatsApp messages analyzed across 70+ profiles with NLP
- 1,514+ Obsidian notes with 898 daily journals spanning 3.5+ years
I track everything because you can't improve what you can't measure.
Easy ████████████████░░░░░░░░ 66 solved
Medium ████████████████████████░░ 130 solved
Hard ████░░░░░░░░░░░░░░░░░░░░ 13 solved
Total: 209 | First-pass AC: 66.5% | Avg submissions/solve: 1.88
| Organization | Role | Impact |
|---|---|---|
| MOSIP | Contributor | Digital identity infrastructure · 6 merged PRs · 3 Shipped Features |
| OpenLake | Coordinator | 60+ repos · CI/CD pipelines · 20+ critical bugs found |
| FOSSEE | Summer Fellow | OSDAG structural analysis |
| skpro | Contributor | Probabilistic regression, CDE, GLMs |
| Component | Tool |
|---|---|
| OS | (i use arch btw) - EndeavorOS |
| Terminal | Kitty |
| Multiplexer | Tmux |
| Editor | Neovim + LSP |
| Knowledge Base | Obsidian (1,514+ notes) |
| Shell | Custom config |
| Version Control | Git + GitHub |
"If you want to understand something, build it from scratch."
·
