Jun – Aug 2026
StimMap3D
Independent project: a browser-based 3D visualization with a live-recomputing electric-field heatmap, shipped with a Methods & Limitations page and a self-check quiz.Illustrative model — not for clinical use
Paige Rattenberry
I build the instruments that check what AI systems claim — evaluation harnesses that run as merge gates, retrieval that traces every claim back to its source, and research on whether a model's explanation points at the evidence. Mostly where being wrong has a real cost.
Currently at Hammerspace, on an internal knowledge platform: an MCP server under access-control filters, an agent loop with citation guardrails, and a scheduled audit job that opens its own remediation pull requests. Primary engineer across 300+pull requests as primary engineer on the Hammerspace knowledge platformConfirmed by Paige on 2026-10-06; updated from the resume of 2026-09-11.Source: Facts confirmed by Paige, 2026-09-08 to 2026-10-06 pull requests. Experience
Built with AI tools; what they got wrong, and which check caught it, is published. How this site was built
Selected work
Apr – Aug 2022
Testing medical-AI explanations
Scored fiveclass activation map (CAM) explanation methods comparedSource: Resume, revised 2026-09-11 explanation methods against ground truth induced by watermarking chest X-rays. FullGrad was the most robust and precise.
Jan – Aug 2022
Measuring scoliosis brace pressure
Originated the project and formed the six-personcapstone team sizeThe team's members are named in SFU's article on the project.Source: Resume, revised 2026-09-11 team. Best Overall Project at ICAMES 2022.
Skills
- LLM evaluation harnesses
- Grounding & claim verification
- Agent loops & tool calling
- MCP
- RAG
- Interpretability (XAI)