Projects
Things I built, and what building them taught me — including the parts that went wrong. Written for someone who wants to know how the decisions were made, not just which technologies ended up on the list.
The platform behind this site
2025 – presentAn LLM gateway with real spend controls, an Avro-typed streaming pipeline, single sign-on, and a host reproducible from one playbook. This site is the visible tenth of it.
FastAPILiteLLMKafkaAvroRedisCelery+5 more
Read the write-upReinforcement learning the exit decision
2026A Q-learning agent that decides, quarter by quarter, when a leveraged buyout should sell — trained live in your browser on a stochastic LBO, with the learned policy, the Bellman updates and the baselines all visible.
TypeScriptQ-learningCanvasMonte Carlo
Read the write-upDeep learning for Korali
2022 · MSc thesis, ETH ZürichA deep-learning module for a high-performance scientific computing framework, verified and benchmarked against cuDNN, oneDNN, PyTorch and TensorFlow.
C++CUDAcuDNNoneDNNPyTorchTensorFlow+2 more
Read the write-upGPU-accelerated Bayesian inference
2021 · Semester project, ETH ZürichMoving the dense linear algebra behind approximate Bayesian inference onto the GPU, and measuring where that actually pays off.
CUDAcuSOLVERC++Rnvprof
Read the write-up