Building well, shipping fast — and a slightly undeniable god complex, someday soon.
- 🛠️ I work across the stack — Go distributed backends, React realtime frontends, and the infra/CI that ships them.
- 🧠 I build LLM agent systems — memory, orchestration, judging pipelines, and cost observability.
- 🔧 Currently building fabri — a local memory + orchestration layer for custom LLM agents (Qdrant-backed tactical/strategic memory, polyglot subprocess tools, agent-as-tool composition).
- 🏆 Smart India Hackathon 2022 winner — NaksheKadam.
- 🎨 Comfortable from schema design to pixels — and the migrations, queues, and deploy pipelines in between.
⚙️ Backend & distributed systems (Go)
- Build correctness-critical services and multi-stage data pipelines that stay sane under failure.
- Make stateful systems reliable — work that's never dropped or double-processed, safe migrations over live data, and databases tuned to survive real load.
🖥️ Frontend (React)
- Architect the data layer, not just the screens — realtime server/client state sync, caching, and patterns a team can build on.
- Ship dense, data-heavy tooling — admin consoles, health dashboards, and views with server-driven search & pagination.
☁️ Infra, CI/CD & releases
- Own the path from commit to production — polyglot monorepos, isolated environments with per-PR previews, and pipelines that catch drift before it ships.
- Gatekeep releases, review others' work, and debug messy realtime/voice systems end to end.
| Project | Description |
|---|---|
| 🤖 fabri | Local memory + orchestration layer for custom LLM agents. Qdrant-backed tactical/strategic memory, polyglot subprocess tools, agent-as-tool composition. |
| 🗺️ NaksheKadam | 🏆 Winning project of Smart India Hackathon 2022 (web version). |
| 🏛️ IoT-Architecture | A curated hub of architectural patterns, principles, and examples for designing scalable, modular, secure systems. |
| 🔐 appwrite_ui_auth | A Flutter library for drop-in template logins backed by Appwrite. |
| 🖼️ Image-Caption | BLIP image captioning + GPT-2 — generates joyful responses to image captions. |



