I work on execution layers, zk systems, and infra that has to handle real load.
zkEVM prover Trying to actually prove EVM execution end-to-end. Not just small circuits — full state transitions, batching, making it practical.
ScanX Indexer for multiple EVM chains. Stores everything — blocks, txs, logs, traces. Built for speed because most indexers are slow or incomplete.
Sentinel Execution proxy layer. Mostly about keeping things modular and not tightly coupled.
Code Intelligence Engine Parses large codebases into structured graphs (calls, flows, dependencies). Lets a model reason over execution instead of raw text. Goal: debugging + deep understanding, not autocomplete.
Low-latency LLM runtime Custom setup for running 7B models locally. Focus on:
- fast inference
- memory efficiency
- streaming + tool usage
Making small models actually usable for real workflows.
Task Execution Engine System where models don’t just suggest — they execute. Handles:
- planning → action → verification loop
- file system + API operations
- failure recovery
Goal: reliable agents, not demo agents.
Go, Rust, TypeScript EVM, Cosmos SDK zk (Groth16, circuits) Distributed systems LLMs (local inference, systems-level integration)
GitHub: https://github.com/Noooblien Email: rsm050501@gmail.com
That’s it.