About
Mehdi Bayat
I'm a senior full-stack engineer based in Vienna, currently building emBET, Sportradar's betting widget embedded inside live sports streams and sportsbook partner sites, handling millions of concurrent users during peak events. Before that: real-time trading and market-data platforms for Iran's national stock exchange, a bank's CRM and CMS platform, and a full rewrite of the logistics system running Iran's 13 main maritime ports.
Most of my career has been in places where latency is measured in milliseconds and downtime costs real money — trading floors, live betting, port logistics. That shapes how I think about software: architecture proportional to what's actually required, trade-offs stated explicitly instead of hidden, and systems built to fail in understandable, isolated ways rather than all at once.
LivePulse, the project featured on this site, reflects that directly. It's built to the same standard I hold production work to: real data instead of a simulator, ADRs that state the actual trade-off instead of the ideal one, and a documented real incident (a WebSocket subscribe race condition) rather than a portfolio that only shows what went right.
On the AI side, I built Sportradar's first production RAG pipeline on AWS Bedrock, and I use AI tools daily for implementation, test generation, refactoring, and code review — but the engineering judgment behind what to build and how to verify it stays mine.
Where latency, availability, and reliability actually mattered
- A stock exchange with a zero-downtime requirement, where race conditions in concurrent WebSocket streams only surfaced under real trading load.
- A live betting platform where a 180ms hot-path latency was a real production problem, not a benchmark exercise.
- A port logistics rewrite that had to stay live across 13 ports with no big-bang cutover.
- A banking CRM where a UI bug wasn't an inconvenience — it blocked someone's job.