ML Engineer - MLOps & Data Infrastructure
I work on production ML platforms: feature pipelines, model serving, and the observability and governance layers around them. Lately I'm spending most of my time on agentic infrastructure - orchestration, cross-agent memory, and guardrails for running autonomous agents safely in production.
- ML platform patterns - multi-tenant model serving, feature stores, governance-by-design
- Data pipelines - orchestration with dbt and Airflow, batch and streaming
- MLOps tooling - CI/CD for ML, model registries, drift and performance monitoring
- Agentic infrastructure - agent orchestration, cross-agent memory sync, prompt-injection guardrails
- fti-continuous-ml - Continuous model training using the FTI (feature/train/inference) framework
- orchestrate-dbt-on-airflow - dbt on Airflow via Astronomer Cosmos, with local DuckDB dev and Feast materialization
- kubernetes-for-ml-engineers - Stand up an ML platform on a laptop/VM, promote it by changing one endpoint
- the-arrow-ecosystem - A self-hosted lab covering the Apache Arrow ecosystem layer by layer - client API (ADBC), wire protocol (Arrow Flight SQL), query engine, and object storage
- strata-ai - SDK for enterprise AI agents, APIs, and ML pipeline contracts
- hermify-mcp - Cross-agent skill and memory sync over MCP
- duka-ai - Simulated, guardrail-first support assistant for a fictional Nairobi electronics shop
AWS · Kubernetes · Terraform · Docker · Kafka · Spark · dbt · PostgreSQL · Feast · LangGraph · PyTorch
I write about ML platform and agent infrastructure design at felixmt.substack.com.