About

Heyyy! I'm a software engineer focused on data, full-stack, and other software "around" AI to make it behave. As most engineers, I am here to build, fix and improve things. I get passionate at work, and excited when I see things work as designed + fix if they're not.

I'm open for roles building full-stack tools around AI systems: the data flywheel, observability, evaluation and model lifecycle. Also interested in physical AI, fleet management and telemetry problems at scale.

Here are some things I can do (also, check out my resume):

Highlights

  • At Netflix, I built tooling for a fleet of millions of agents and ML models powering recommendations and GenAI workflows. I worked on analytics and operations, on a lot of full-stack internal apps and Kafka for online model observability (detecting drift, null inflation, missing data). I unified 3 ML logging libraries into one path, which saved about $5 million a year.
  • At SpaceX / Starlink, I led a small team of rockstars for fleet management software. We helped scale customers 100x (50k → 5M+) and satellites/other ground assets 20x. We used gRPC, Python, C++ and C#. The tools supported hundreds of operators receiving thousands of alerts a day and sifting through gigabytes of telemetry per second. The app we built is still key for safe and reliable 24/7 operations of the Starlink fleet.
  • At CERN, I helped with Large Hadron Collider (LHC) reliability by developing test software and electronics for power converter controls.
  • Additionally, I have experience with electronics, robotics, and space. I can blink the ◉ LED, debug a PCB, or program an FPGA.

Stack

  • Web / services: Python (Flask, Django, FastAPI), Go, Java, Scala, C# / dotnet, C++, TypeScript, Angular, React. APIs: REST, GraphQL, protobuf, grpc.
  • Data / ML platform: Metaflow, MLflow, Sagemaker, Kafka, Flink, Spark / Spark SQL, Parquet, Airflow, Databricks, Docker / k8s, AWS (S3, EC2, Athena), PostgreSQL, MongoDB, CockroachDB, Elasticsearch, Datomic.
  • AI: pandas, tensorflow / keras, scikit-learn, mainly for supervised learning. Viz: plotly, DASH, Streamlit, Bokeh, D3. Evals, feature drift, feature store. Arize, Braintrust, OpenTelemetry collectors / SDKs.
  • Embedded: connected sensor networks, embedded ↔ cloud interfaces, embedded C / C++, bootloader (U-Boot) and OTA, basic VHDL/Verilog on Xilinx and Altera (SoC drivers and DMAs), oscilloscope, logic analyzer, and PCB debug.

... and some more. I'm used to wearing many hats, and just doing what's needed at the moment, and learning on the job. Speaking about that: I ❤ learning new things, as well as teaching and talking about them.

At work I can speak English, Polish and ~French. After work, I can (sometimes) succeed at communicating in basic German, Italian, Russian. I am currently learning Chinese and JavaScript.