The portrait is regenerated from my GitHub avatar on every run, and the stats refresh daily through a pinned GitHub Action.
I am Anand B - Co-Founder at Arbor, computer science student at VIT Vellore, and a systems builder obsessed with making complex software legible.
My main bet is Arbor: graph-native code intelligence that maps a repository with ASTs, walks the real call paths behind a pull request, and gives humans and coding agents the exact blast radius before merge.
The same commit should produce the same answer. Deterministic context beats plausible guesswork.
|
The map AI coding agents are missing. A Rust graph engine, PR blast-radius system, and MCP context layer built around deterministic program understanding instead of embedding-based RAG.
|
Ground truth for APIs used by coding agents. Blocks the APIs that do not exist and serves the ones that do - one index, queried in both directions.
|
|
Structural weight paging for consumer GPUs. Experiments in local adapter routing, just-in-time model state, and the telemetry needed to know when the system is actually winning.
|
A safer, bloat-free Lenovo Vantage replacement. A Windows-first control plane with read-only discovery, explicit capability evidence, and a short-lived elevated broker for guarded hardware access.
|
More systems from the lab
- Runoscope - the portable HAR file for AI coding-agent runs.
- Agent Lens 2.0 - visual debugging for agent workflows.
- Zubaan - live multilingual compliance checks for Indian financial sales.
- FinShield - financial document forensics and graph investigation.
- witr - process, port, and service causality diagnostics.
I optimize for determinism over vibes, local-first over rented black boxes, instrumentation before optimization, and safety gates before privileged writes.
Languages: Rust Python C# / .NET TypeScript Dart
Systems: AST parsing code graphs MCP agent observability local AI Windows internals
Default posture: build deeply -> measure honestly -> ship relentlessly