AI Infrastructure Builder & Open Source Maintainer
I leverage AI to build high-complexity, local-first infrastructure tools. My focus is on solving the Context Window Bottleneck in developer workflows through semantic compression and distributed system design.
π Architect of Vidurai
I am the creator and lead maintainer of Vidurai (v2.0.0), a local context middleware ecosystem.
I designed Vidurai to act as an abstraction layer between high-velocity developer telemetry (IDE, Terminal) and LLM context windows. It solves Information Entropy problems by treating developer context as a graph, not a log.
System Architecture Designed:
- Distributed Topology: Orchestrated a Hub-and-Spoke architecture connecting VS Code, Chrome, and Python daemons via WebSockets.
- Concurrency: Implemented non-blocking event ingestion using Python's
asyncioand SQLite WAL mode. - Deterministic Logic: Designed the SF-V2 Algorithm to semantically compress code history, reducing token overhead by ~70%.
I build with a focus on latency, local-first principles, and modularity.
- Core: Python 3.9+ (FastAPI, AsyncIO), TypeScript
- Data: SQLite (Advanced configurations), JSONL
- Transport: WebSockets, HTTP/JSON-RPC, IPC
- DevOps: Docker, GitHub Actions, PyPI Automation
- vidurai: The core SDK and Compression Engine.
- vidurai-daemon: The local background service (FastAPI).
- vidurai-vscode-extension: The telemetry agent for VS Code.