Principal Software Engineer — backend systems and relational database performance.
I own the price and promotion domain at Digikala, the largest e-commerce platform in its market. Six years of making systems cheaper, faster and more correct while they stay in production — usually by arguing for the change nobody wanted to own.
My specialism is the database layer: the point where a schema decision, an index, or an isolation level quietly decides what the whole platform can afford.
100M+ events a day through buy-box selection, at 2,000+ req/s on five database CPU cores
8B+ rows in our largest table, split into hot and cold halves that both serve live traffic
→ hot table 32× smaller, zero downtime
40% less CPU on the main database, after tracing a "caching bug" to the transaction
isolation level across 10+ live databases
50k+ req/s stock-management microservice with distributed transactions
90→10 minutes for auto-pricing across 1M+ products
800k concurrent users across three Black Friday campaigns
~200 engineers through an advanced MySQL programme I designed and taught
Languages Java (Spring Boot, Spring Data JPA, Hibernate) · PHP · Python · Go · C#/.NET (ASP.NET Core, .NET Framework, Entity Framework) · JavaScript with Node.js and React · C · R · SQL · Bash
Databases MySQL and PostgreSQL performance — query optimisation · index design · transaction isolation and locking · replication, failover and disaster recovery · ClickHouse · MongoDB · Elasticsearch · IoTDB/InfluxDB for time series
Caching & messaging Redis · Memcached · write-around and write-through caching · Apache Kafka · change data capture · REST API design · OAuth 2.0 and JWT · MQTT · WebSocket
Architecture Microservices · distributed transactions · CQRS · eventual consistency · SOLID · clean architecture · domain-driven design principles · hexagonal (ports-and-adapters) separation · vertical slice organisation
Testing TDD · unit, integration, contract and load testing, including write load tests run against production
AI Claude Code and Cursor daily on a production codebase · LLM features and agentic automation in production workflows
Tooling Docker · Linux · Git · GitLab CI/CD · Maven · Grafana · Percona Monitoring and Management · Jenkins (familiarity)
Most of what I build is closed-source and lives inside a marketplace. What is public here is mainly systems and database work from research and university projects — the fundamentals underneath the production work.
| Project | Built with | |
|---|---|---|
| mariadb-tpcds | C, Python | The full TPC-DS v4.0.0 schema and all 99 queries against MariaDB, with an isolated measurement protocol. Benchmarking done carefully enough to be honest about what the numbers do and do not mean. |
| encoding-compression-benchmark | Java, Python | Benchmarks comparing numeric encoders, compressors and hybrid pipelines, with analysis scripts. Came out of asking what it actually costs to store a time series. |
| kafka | Python, Java, Docker | Server for a Kafka-style message broker — brokers, partitions and delivery |
| onlineTicket | JavaScript, Docker | An online ticketing platform with organisations, products and agent assignment, shipped through a full CI pipeline. |
| mibench | C, Python, Java | The MiBench embedded suite extended into Python and Java, plus the gem5 → McPAT toolchain for power, area and timing. The measurement infrastructure behind a thermal-management paper. |
| c-minus-compiler | Python | A one-pass compiler: scanner, parser, semantic analyser and intermediate code generation, including recursive calls. |
| OS_pintos | C | PintOS through three phases, ending in an inode-based filesystem with a buffer cache and the synchronisation to make it safe. |
| MIPS-Verilog | SystemVerilog, Assembly | A MIPS processor — the other end of the stack from a query planner. |
| SDN | Python, Mininet | Shortest-path routing pushed into a software-defined network controller over a Mininet topology. |
Lightweight MQTT–Time Series Database Integration for IoT: Optimizing Payloads for Traffic and Energy Efficiency — profiling Apache IoTDB in a clustered environment and designing an MQTT payload format that cuts network and energy overhead. In preparation.
HAMLET: Adaptive Machine Learning-based Thermal Management in Fault-Tolerant Heterogeneous Multicore Systems — with A. Ansari and M. Ansari. Submitted.
LinkedIn · h.aghaei.araei@gmail.com