Zk-rollup DEX
Backend and on-chain work. Canonical state logic in Rust, proofs generated with SP1, settlement through Solidity contracts on Ethereum.
RustSP1Solidity
Data platforms / trading systems
I build and run backends for crypto trading and analytics. That has meant ingestion pipelines at billions of rows, APIs with sub-100 ms p95, and leading the team behind them. Today I'm building exchange and zk-rollup infrastructure.
Pune, India · Open to remote & relocation
Backend and on-chain work. Canonical state logic in Rust, proofs generated with SP1, settlement through Solidity contracts on Ethereum.
Built the Go backend: order matching engine, risk management system, durable balance ledger, a price oracle built from major exchange prices, and an in-house market maker.
Built the Python retrieval and orchestration layer for an exchange's support and marketing assistant, using RAG and hybrid search over ClickHouse.
Chain data streams over WebSockets into Kafka. PostgreSQL is the source of truth, ClickHouse serves analytics and Elasticsearch serves search. Writes are idempotent, with retries, backfills and reconciliation, so missed data can be replayed safely. Background jobs run on Celery and RabbitMQ.
Heavy P&L reads caused a production outage. Rolling P&L on a 2B+ row Solana trades table stayed slow for wallets with millions of trades, even with indexes and per-token batching. I moved P&L analytics from PostgreSQL to ClickHouse.
Repeated P&L requests for one high-volume wallet overloaded the database and stopped indexing. I killed the running queries, disabled those reads, and limited P&L to wallets under a trade-count threshold using a precomputed trade-count table.
Owned the Python backend serving 10,000+ API users and 100,000+ monthly active users at peak, at 500,000+ requests a day.
Sign-in with Google, JWT sessions, hashed API keys, role-based access per subscription tier, and token-bucket rate limiting per user and per IP.
Managed the backend and DevOps team. I planned and assigned work, ran design and code reviews, and was the technical contact for product and the client.