Re:Ask HN: Who wants to be hired? (October 2026)
by vic_ml_throwawa on Thursday October 15, 2026 @02:04AM
(#50099817)
Location: Victoria, BC, Canada
Remote: Yes
Willing to relocate: No
Technologies: PyTorch, Python, C++, 3D medical image segmentation, AWS, LLM/agent systems (Anthropic API, MCP)
Résumé/CV: on request
Email: [email protected]
Staff-level ML / computer vision engineer, currently at a medical device company. My work lives where ML meets regulation:
- Designed from scratch a 3D deep-learning segmentation model for ultrasound surgical navigation (multimodal transformer, ~2.1M params): Dice 0.86, median HD95 4.5mm on a frozen held-out set — at a fraction of the training data published benchmarks used.
- Deploy imaging-based cancer treatment prediction on AWS, including the technical documentation for its regulatory submission.
- Built the evaluation discipline around it: 3D Hausdorff/surface-distance frameworks, bootstrap scaling analysis projecting data needs for regulatory-grade validation, Monte Carlo error propagation through the full segmentation→registration→tracked-tool chain.
- Before ML: 3 years of hard real-time embedded C++ on QNX for Class C devices (IEC 62304, MISRA C++:2023, ISO 14971). I validate like lives depend on it, because they do.
- Production agentic LLM systems: multi-agent MISRA compliance automation and regulatory doc traceability. Ran a controlled study of LLM-assisted unit test generation — 13.75 min saved per test, ~23 hrs per 100, gated on six quality criteria. The point wasn't speed; it was knowing exactly when the output is trustworthy.
Looking for: short-term contracts in medical imaging ML, computer vision, or LLM automation for engineering teams. $135/hr, ~10 hrs/week alongside my full-time role, remote only.
One honest boundary: I haven't shipped edge inference optimization, so I won't bid on it.
Staff-level ML / computer vision engineer, currently at a medical device company. My work lives where ML meets regulation:
- Designed from scratch a 3D deep-learning segmentation model for ultrasound surgical navigation (multimodal transformer, ~2.1M params): Dice 0.86, median HD95 4.5mm on a frozen held-out set — at a fraction of the training data published benchmarks used. - Deploy imaging-based cancer treatment prediction on AWS, including the technical documentation for its regulatory submission. - Built the evaluation discipline around it: 3D Hausdorff/surface-distance frameworks, bootstrap scaling analysis projecting data needs for regulatory-grade validation, Monte Carlo error propagation through the full segmentation→registration→tracked-tool chain. - Before ML: 3 years of hard real-time embedded C++ on QNX for Class C devices (IEC 62304, MISRA C++:2023, ISO 14971). I validate like lives depend on it, because they do. - Production agentic LLM systems: multi-agent MISRA compliance automation and regulatory doc traceability. Ran a controlled study of LLM-assisted unit test generation — 13.75 min saved per test, ~23 hrs per 100, gated on six quality criteria. The point wasn't speed; it was knowing exactly when the output is trustworthy.
Looking for: short-term contracts in medical imaging ML, computer vision, or LLM automation for engineering teams. $135/hr, ~10 hrs/week alongside my full-time role, remote only.
One honest boundary: I haven't shipped edge inference optimization, so I won't bid on it.