Milestones, publications, and updates from the Krv team.
We're building a new Erdos and Renyi product line for clinical trial intelligence and patient recruitment that runs fully on a site's own machines. The names are inspired by the mathematicians Paul Erdős and Alfréd Rényi, and by their pioneering work on random graphs. Patient records never leave the site. Updates coming soon.
Ten minor releases since v0.4.0 in July. Scores are now anchored to quality gates with advisories relative to your own codebase, Topos ships as a portable Agent Plugins 1.0 package, and v0.14.0 shares dependency graphs and reuses analysis so repeated MCP assessments run faster on less memory.
Krv co-founder Jeremy Wayland defended his TUM dissertation, "Understanding Data Representations using Geometry and Topology", and was awarded magna cum laude.
Phirewall is a local PHI vault and MCP gateway: AI assistants can analyze patient records while names, record numbers and birth dates stay on the clinician's machine as codes. It fails closed and has no servers of its own. Coming soon for Mac and Windows.
Krv Labs is now Krv Labs, Inc., a Delaware C-corporation.
The team spent the summer working out of Madiba's offices in Lake Forest, California. Thanks to Madiba for hosting us.
Topos v0.4.0 moves the engine to Rust, lists Topos on the MCP Registry, and welcomes its first outside contributors.
Quality is the new currency for agent-written code. Topos v0.3.0 is the first proper release of open-source structural quality metrics for coding agents—scoring whether diffs are built to last, not just whether they pass tests. Evaluates whether programs are SIMPLE, COMPOSABLE, and SECURE. Currently supports code written in Python, Rust, C++, TypeScript, and JavaScript.
Conversations at the International Pediatric and Lifespan Data Science Conference, after our podium talk, led to our first work with a pharmaceutical sponsor. We model a trial's target patients, and each site finds candidates in its own records, locally. Investigators review every candidate; Krv never sees the patients.
Krv was selected as one of three winners of the top abstract submissions and gave a podium talk on Pasteur, our clinical evaluation product, at the 3rd Annual International Pediatric and Lifespan Data Science Conference.
Krv Labs and Compass Health partnered with Women in Tech & Entrepreneurship (WTE) and the USF Engineering Council to host the 'Deployment Dreams' hackathon. Participants built robust sleep stage classification models designed to handle real-world sensor corruptions in wearable data.
Krv Labs has been accepted into NVIDIA Inception, a program designed to help startups accelerate their AI solutions and growth with access to NVIDIA resources, technology, and go-to-market support.
J. Wayland submits cumulative dissertation titled: "Understanding Data Representations using Geometry and Topology"
Website updated to showcase Krv's focus on improving evaluations and reliability for clinical AI deployments.
Krv built and launched the Compass Health platform for care coordination.
J. Wayland and team won first place in the Novartis Challenge by developing a geometric deep learning pipeline for drug-target interaction prediction.
The article discusses a study that classifies U.S. coal-fired plants into typologies using Krv's THEMA framework.
Research featured on AAAS's official press release platform, highlighting data-driven framework for faster U.S. coal phaseout.
Peer-reviewed paper on scalable risk valuation for energy infrastructure.
The team released Thema v0.1.3 and Retire v0.1.1—open-source tools that support research to accelerate the coal phaseout.
Krv team and collaborators are notified of acceptance decision.