AI Architect @ Dell Technologies · Founder @ AroorA AI Labs · 18+ years building enterprise systems
I build open infrastructure for trustworthy agentic AI. One principle behind everything here: AI systems should be verifiable, not just impressive.
| Project | What it does | Install |
|---|---|---|
| ISON | Token-efficient data format for LLMs — the JSON alternative for AI workflows | pip install ison-py |
| ISONGraph | Property graphs for LLM context — ~70% token savings, 92% traversal accuracy | pip install isongraph |
| MAPLE | Multi-agent protocol engine — typed messages, resource negotiation, Result<T,E> error handling |
pip install maple-oss |
| RudraDB | Relationship-aware vector + knowledge-graph database | pip install rudradb-opin |
| Contexel | Deterministic context-economy stages for code-writing agents — dedupe, relevance scoring, token budgets | pip install contexel |
| RudraMem | Deterministic agentic memory — 85.4% LongMemEval, 13× faster ingestion, zero LLM extraction calls | coming soon |
| SnapLLM | Multi-model serving — sub-ms model switching, 20B inference on CPU at ~38 tok/s | — |
In the lab: UniPass (forward-only training) · ZiZu (governance-first agent platform) · Radically (distributed multi-modal inference)
ISON 
11 publications on AI governance, privacy-preserving ML, agent memory, and multi-agent systems — Google Scholar
LinkedIn · rudradb.com · ison.dev · graph.ison.dev
