about

Aditya Pandey

Rust engineer. I work on execution and consensus node infrastructure, databases, and AI systems. Shipped real-time distributed systems and several DeFi protocols along the way. The closer I look at something, the more I want to change it.

That includes reth — a modular, high-performance execution client. Work merged across sparse trie optimization, engine tree performance, and pipeline correctness.

Currently

  • ›$30.18 mean-edge market maker for Paradigm's Prediction Market Challenge — built around a closed-form volatility gate and monopoly-quoting regime
  • ›Active contributor to paradigmxyz/reth — sparse trie optimization, engine tree throughput, and execution pipeline correctness
  • ›tiny-lsm: custom database engine in Rust, extending the Mini-LSM foundation with time-window compaction
  • ›system-eye: per-app resource policy governor for macOS with an MCP layer for AI agent integration
  • ›Exploring LLM inference infrastructure — serving latency, KV cache management, and batching strategies
  • ›VerseFi: cross-collateral structured products for fragmented prediction markets (Solana Colosseum hackathon)

Open Source

mergedclosed / under review

paradigmxyz/reth →
reth-deploy

Provisions im4gn NVMe-backed spot instances, compiles reth with jemalloc-prof in the background, hot-swaps the binary into the running container, and exposes samply CPU + pprof heap endpoints — used for local perf investigation before submitting upstream.

perf(reth-engine-tree): sparse trie bulk move new storage updateFeb 2026fix(pipeline): prevent unwind beyond history limitsJun 2025fix: drop support for total difficulty tableOct 2025feat(trie): add assert_eq_parallel_sparse_trie_proof_nodesJun 2025perf(storage): merge-scan HashedStorages writes in write_hashed_stateFeb 2026perf(engine): pre-populate keccak cache before parallel tasks beginDec 2025
all contributions →
tempoxyz/tempo →

Profiling-driven contributions to Tempo's transaction pool and payload builder — diagnosing write path lock contention, prewarm pipeline inefficiencies, and pool maintenance hot paths that surface only under production transaction throughput.

all contributions →

Stack

languages
RustTypeScriptSolidityGo
systems
eBPFMDBXtokiorethalloy
databases
LSM treesRocksDBPostgreSQLRedis
AI / infra
MCPLLM inferencePrometheusGrafana

Contact

githubi-m-aditya
x / twitter@_0xsensei_
linkedinadityapandey1997
resumeview pdf →