Predictable systems
Reliable, observable, and maintainable.
About Bahman
I build boring automation
for expensive problems.
I'm Bahman Shadmehr, an independent AI systems and automation engineer with nine years of production software engineering experience. I help teams turn messy workflows into dependable systems they can understand, operate, and improve.
01 — Work so far
Built and run in production.
Since 2017 I have built and run production systems as a backend and infrastructure engineer. That includes a payment gateway's move from a monolith to microservices on Kubernetes, store integrations for e-commerce platforms, and location-tracking backends for logistics. It also includes automated trading pipelines for U.S. energy and stock markets, each with its own monitoring and alerting.
Most of it was Python (Django, Flask, Tornado, FastAPI, Celery) on Docker and Kubernetes. On AWS that meant S3, Lambda, Step Functions, ECS, DynamoDB, Athena and Glue. On GCP it meant container batch jobs. The work shipped through GitLab CI or GitHub Actions and was watched with Prometheus, Grafana, Sentry or ELK. That background is behind the controls on this site: durable records, retries that do not repeat side effects, and someone who can see when a job fails.
Since 2025 I have worked independently. For one client team running custom services on Kubernetes, I built an alert-analysis system: a routing layer decides which logs, tools and runbooks to consult, how many API calls to make and which model to use, and a cache answers repeated alerts without a new analysis. It explains what went wrong and how it should be fixed; an engineer decides what to do. The team still uses it.
For another, I built behaviour analysis on top of rrweb session recordings: AI-generated funnels, a filter that decides which sessions are worth a full analysis, and reuse of earlier results when a user's behaviour matches one already analysed. It is still in use.
02 — My commitment to you
Principles over promises.
I won't sell you magic. I'll help you understand the real problem, design the smallest reliable system that solves it, and leave you with something your team can operate.
The design studies show how I decide; the open-source packages show how I build; pricing shows what a first step costs.
Visible uncertainty
Clear about what is known, inferred, and unknown.
Human decisions
Automation supports judgment. It doesn't hide it.
Code, tests, operating notes
A system your team can run after handoff.
04 — Patience in practice
Fishing is where I stop rushing.
I enjoy the preparation, the waiting, and the attention it demands. Not every trip ends with a catch. That is part of what makes it worthwhile.
A real catch. A good day outside.
05 — A wider perspective
Zoom out. Build for the long term.
I care about systems that still make sense after the first release: maintainable code, visible trade-offs, and room to change when reality changes.
07 — Let's build what's next
Real problems. Reliable systems.
If you're working through a costly workflow, fragile automation, or an AI system that needs stronger foundations, I'd like to hear about it.
Same curiosity.
Different terrain.