Senior Software Engineer · ConsenSys
Jun 2021 – PresentLondon, UK
Built ConsenSys Staking from the ground up — the Ethereum staking backend behind MetaMask, serving institutional clients at 33k+ validators and $2B+ staked.
- Established SDLC practices as an IC without formal authority: OpenTelemetry-by-default, RFC-first features, ADRs for decisions, repo standards, and CI gates — adopted across my team and picked up by adjacent teams.
- Designed a multi-tenant API prioritizing correctness over throughput: row-level-security isolation, idempotent writes, audit trails, and a checkpointed state machine — no duplicate side-effects, safe resume after partial failures.
- Lighthouse: fixed an O(n) → O(1) keystore-decryption path causing validator restarts; upgrades resumed at 99.9% attestation. Issue #4936 → PR #4126.
- Built and operated a production MEV-Boost fork for a tier-1 exchange: relays configurable per validator, hot-reloadable (atomic, lock-free swap), with the Prometheus instrumentation it originally lacked — ~26% lower latency. Issue #455.
- Designed and built a Kubernetes snapshot operator for Ethereum nodes, replacing week-long resyncs with snapshots to cut cluster rebuilds from days to minutes and enable rapid provisioning and disaster recovery.
Senior Software Engineer · Clearmatics
Aug 2019 – Jun 2021London, UK
Proved BFT proof-of-stake viable on an EVM chain — a working prototype, before Ethereum's Merge.
- Patched go-ethereum to run a Tendermint-style BFT consensus layer.
- Cut build times ~40% — rebuilt CI with caching, parallelization, and reproducible artifacts.
- Hardened the consensus paths with characterization tests + TDD, raising coverage to ~80% so the trickiest code was safe to change.
Head of Research & Development · DAO.Casino
Oct 2018 – Aug 2019Moscow, Russia
Proved verifiable, unbiasable on-chain randomness feasible — end-to-end PoC on Tendermint/Cosmos.
- Built a threshold-BLS random beacon into Tendermint consensus: the proposer combines validators' precommit shares into the block's random value — but must commit the block's transactions before that value exists, so it can't predict or bias the outcome.
- Unbiasable and publicly verifiable — strong enough for a game of chance, not just cryptographic sortition. Delivered the full prototype: patched Tendermint, a Cosmos SDK app (randapp), and a kyber-based Go DKG library, with tests validating it end-to-end.
Tech Lead, Core Backend · Humaniq
Nov 2017 – Oct 2018Moscow, Russia
Mobile-banking app for the unbanked — Go service platform across ~20 services, ~10-person team.
- Brought a fragmented estate of ~20 services under one standard: a Go service template + shared libraries (gRPC/HTTP, structured logging, tracing, panic recovery) and clean-architecture/DDD conventions that a team of ~10 shipped every new service on by default.
- Owned the platform’s hardest service hands-on — a wallet moving user funds: every transfer gated on multi-step validation of both sender and receiver (geo, blacklist, account status) before async settlement via RabbitMQ.
Senior Software Engineer · Lazada (acquired by Alibaba Group)
May 2015 – Nov 2017Moscow, Russia
Core product-catalog backend — 16k+ RPS/instance, p95 <100 ms, 99.9% through 11.11 / 12.12 surges.
- Data access & persistence: DAL with CQRS read/write separation — writes to masters, reads from replicas across sharded/partitioned MySQL; read latency p99 <120 ms while isolating write load.
- Microservices infra (pre-Kubernetes): etcd service discovery + Docker orchestration for reliable rollouts and fast instance replacement; centralized Elastic Stack logs and Zipkin-style tracing for on-call.
- Legacy PHP → Go migration: incremental replacements behind a compatibility layer with production-example tests — lower latency/CPU per request, safe staged cutovers.
Head of Development · Auction & Tender House
Mar 2012 – May 2015Moscow, Russia
Regulated e-procurement platform — FSB/FSTEK-audited, DDoS-hardened.
- Regulated e-procurement overhaul: rebuilt auction/tender flows with DDD, CQRS, and event sourcing; read/write split + caching cut critical paths from minutes to milliseconds, with deterministic, verifiable records that reduced disputes.
- Availability & resilience: single server → clustered/redundant with upstream DDoS scrubbing and a reserve data center after a ~2 Gbps attack — ad-hoc response replaced by repeatable playbooks.
- Security & compliance: passed FSB/FSTEK-accredited audits for Commercial Secret data; four-eyes release control and immutable audit logs enabled forensic replay and dispute resolution.