At IMU Biosciences, I designed and built a data platform and catalogue for 100k+ single-cell samples, making ML training datasets reproducible. I also developed QC monitoring and synthetic data tools to validate pipelines and models.
At Mobkoi, I rebuilt data pipelines on GCP and led the move to impression-level ingestion, enabling more granular analytics. I now build open-source ML and LLM tooling, including an activation-probing toolkit and a knowledge-graph extraction pipeline.

