Summary
Backend engineer who owns financially critical systems end to end. Technical owner of all four services behind a crypto exchange's futures market-making platform — built from zero to roughly 50 markets, placing millions of orders a day against multi-million-dollar balances at 99.9% uptime. Works AI-first: architecture and planning done in dialogue with LLMs, implementation generated against detailed plans, every line reviewed because the money is real. Computer Science at Sharif, now an MBA in Strategy — comfortable owning the product and business call as well as the technical one.
Experience
- Platform ownership — Own all four services behind the exchange's futures market-making platform, taking coverage from zero to ~50 markets at 99.9% uptime. The system places millions of orders per day against multi-million-dollar balances, where a single error is a direct financial loss.
- Hedge engine — Built and own the hedging engine that mirrors positions across external exchanges for ~25 markets with exact accounting reconciliation, letting the desk market-make those books with no directional risk.
- Risk manager — Proposed running the remaining ~25 markets unhedged under algorithmic risk control instead of hedging every position, then designed and built the service that executes it — generating several thousand dollars of profit within five days from one of the platform's thinnest markets alone.
- Price service — Engineered a sharded streaming service consuming ~30 concurrent WebSocket feeds across three exchanges with REST fallback and per-shard health checks, keeping order-book and mark-price data fresh for every downstream consumer.
- Order placement — Designed the depth and spread algorithms that quote near-zero spreads across all covered markets.
- Incident leadership — Led response to a mispricing event that caused erroneous liquidations: stabilized the affected markets, made every affected user whole, then shipped a pre-deploy sanity-check service, hardened price validation, and a single source of truth for symbol metadata — with no recurrence since.
- Reconciliation — Traced recurring account discrepancies through every price-mutation source, then automated the fix with reconciliation jobs, balance-mismatch alerts, and daily reports.
- Team leadership — Serve as deputy to the tech lead, owning architecture decisions, code review, and on-call; onboarded, mentored, and managed the task flow of two junior engineers.
- AI-first delivery — Drove the team's AI-first workflow: brainstorm architecture and trade-offs with Claude, generate implementations against detailed plans, and review every line given the financial stakes — cutting delivery time by roughly 60%. Authored the shared AI guideline files (prompts and code style) now shipped in every team repository.
- Product ownership — Operate as a technical product manager alongside engineering: set market-maker strategy and represent the team in board and CX meetings.
- Built backend services for Iran's national taxpayer platform, a government system holding detailed records covering essentially the entire population of the country.
- Implemented PKCS#11-based cryptography for sensitive operations — difficult to achieve with the Java tooling available at the time — unblocking a requirement the team had been unable to satisfy.
- Worked across Java, Spring, Kafka event processing, Oracle Database, and Docker.
- Owned the backend of a pharmacy management system, modeling prescriptions, insurance, medications, and patients along with the administrative rules real Iranian pharmacies operate under.
- Delivered REST APIs on Django and PostgreSQL within a four-engineer team, working fully remotely.
Selected Projects
Plain-text task, time, and expense tracker for the terminal, published on PyPI and used daily. Natural-language capture through a pluggable LLM layer (local Ollama, OpenAI-compatible, or Anthropic) that degrades gracefully to regex parsing when no model is reachable. Zero runtime dependencies, human-readable Markdown storage, Jalali calendar support, MIT licensed.
Matching engine microservice with deterministic price–time priority. Runs matching single-threaded to rule out race conditions and double spending, holds the live book in Redis, and guarantees exact crash recovery by replaying a PostgreSQL event log over periodic snapshots. Ships concurrency and recovery test suites, Locust load tests, and OpenAPI documentation.
Authentication service in Go: JWT issuing and validation, bcrypt password handling, Redis-backed sessions, and PostgreSQL persistence through GORM.
Streaming and batch pipeline implementing the lambda architecture across a Spark cluster (master plus three workers), Kafka ingestion, a Cassandra serving layer, and Zeppelin notebooks.
Two Spring Boot microservices that generate and count short URLs, communicating asynchronously over Kafka.
Education
In progress, pursued alongside full-time engineering work.
- Coursework: Data Structures, Algorithms, Operating Systems, Data Engineering, Web Programming, Advanced Programming.
- Completed while working full-time as a backend engineer; organized university camps and events and led volunteer initiatives.