Loading...
Hi, I'm
Mahdi Akbari Zarkesh
Technical Lead, Agentic AI · Product-Minded Engineer
I design, code and ship agentic AI systems end to end: agent memory, RAG with real citations, tool calling, guardrails, and the evals that decide when a change is good enough to ship. I lead the architecture of a production AI agent platform at Alien Intellect, and I grew its engineering team from 2 to 11 in a year. Before that, five years at Cafe Bazaar architecting services at 10,000+ requests per second for 51M+ users. I start from the why and the what, then build the how.
Now: leading the AI agent platform at Alien Intellect
The numbers behind the story
10K+
Requests/Second
Peak load on the update service I architected
51M+
User Platform
Users served by the systems I owned at Cafe Bazaar
2→11
Engineers in a Year
Hired, onboarded and coached the Alien Intellect team
#1
Class Rank
Ranked 1st of 200, Ferdowsi University • M.Sc. from Sharif
What I work with
Agentic AI in production first, the backend and infrastructure that keep it up second, and the product judgment to decide what gets built at all.
Agentic AI Systems
- AI Agents: tool calling, agent memory, orchestration
- RAG pipelines & vector databases
- Prompt & context engineering
- LLM evaluation: LLM-as-judge, regression suites
- AI guardrails & agent safety
- LangGraph, Claude Agent SDK, MCP
Product & Team Leadership
- Technical product management, 0→1 to beta
- Product strategy & KPI-driven prioritization
- Systems & structured thinking
- Team building: hiring rubric, interview flow, levels
- Onboarding & coaching engineers
- Delivery culture: sprint cadence, Definition of Done, QA flow
Backend Engineering
- Python (FastAPI, Django) and Go in production
- Microservices & distributed systems
- gRPC & REST API design
- System design & architecture reviews
- Python-to-Go migrations with zero downtime
- High-throughput services: 10,000+ RPS, sub-70ms p99
Cloud & Infrastructure
- Kubernetes orchestration & autoscaling
- Docker & container sandboxes for agents
- AWS & Google Cloud
- CI/CD pipelines (GitLab, GitHub Actions)
- Observability: Prometheus, tracing, SLOs
- Performance tuning & capacity planning
Data & Storage
- PostgreSQL at ~1TB: indexing & caching strategy
- Redis caching, in-process caches for latency
- Vector databases & embeddings
- Elasticsearch & full-text search
- Apache Kafka & event streaming
- Data modelling for read-heavy services
Research & Teaching
- 3 published papers (IEEE, Springer, Scientia Iranica)
- Blockchain frameworks for healthcare IoT
- Fog & edge computing, workflow scheduling
- Conference talks: Sharif Job Fair, PyDay Iran
- Quera bootcamp instructor: Go and Python
- Engineering writing read by 6,500+ people
Read along
Notes on production AI agents, system design, and the trade-offs behind them, in English and Persian
Building agents that have to work in production?
I like comparing notes on the hard parts: agent memory, RAG that cites its sources, evals that gate a release, guardrails, and the team habits that keep all of it shipping. Read what I have written, or say hello.
✓ 10K+ RPS systems✓ 51M+ users at Cafe Bazaar✓ Team grown 2 → 11