News
- Aug 2025Lab page launched at labs.coderin90.com.
- Jun 2025Joined the Mantis project at MIT CSAIL.
Research
Human-in-the-loop clinical AI
Interactive systems where the clinician stays in command: models that defer gracefully, communicate uncertainty, and explain themselves through evidence rather than assertion. I study when to trust a recommendation, when to override it, and how to measure both.
Safety and evaluation of medical language models
Evaluation methodology for LLMs in clinical work: documentation, summarization, and literature synthesis. Failure taxonomies, benchmarks grounded in real workflows, and guardrails that hold up under distribution shift rather than under lab conditions.
Cognitive cartography for medical knowledge
With the Mantis project at MIT CSAIL: joint multimodal representations that project clinical evidence, literature, and patient data into navigable maps, so humans and AI reason together over one shared, source-traceable picture of the evidence.
Projects
tikkrtrytikkr.com
Live translation for lectures and meetings. Everyone in the room follows in their own language, as captions or as translated audio, and afterwards works from a recording with per-language notes they can search and revisit. Built for institutions and instructors teaching across languages. It is also where my applied research sits: speaker diarization, education technology, and the didactics of learning in a second language.
RegnaVerktregnaverkt.com
An AI data platform for investment companies: one API and MCP server covering financials, filings, ownership, and news for listed and private companies across the Nordics, the UK, and the US.
Coder in 90 Classroomcoderin90.com
A GitHub-native replacement for GitHub Classroom, which shut down in August 2026. Assignments go out as repositories and are auto-graded on every push; when a test fails, the grader reads the real test output and explains the failure and the fix instead of returning a bare score. Behind the feedback is a RAG pipeline that retrieves the assignment brief, the course material, and the student's own code, so the explanation is grounded in what the class has actually covered. That makes a richer experience for students and teachers alike: students get feedback that reads like a tutor who knows the course, not a generic hint, and teachers see where a class is getting stuck without reading every submission. Free for students.
Teaching
I teach as an Engineering Instructor at the Digital Career Institute, where I coach career-changers through full-stack software engineering.
Bio
I am a consultant at Codurance with a special interest in AI and RAG deployments: taking retrieval-augmented systems from a promising prototype to something a team can run in production. That means the unglamorous parts: retrieval quality, evaluation that catches regressions before users do, cost and latency budgets, and answers that show their sources.
My research is on assisted AI in medicine: systems designed to keep clinicians, not algorithms, in command. How models defer, how they show their evidence, and how we evaluate them to a standard a regulator and a physician can both trust.
I am affiliated with the University of Colorado Boulder and with the Mantis project at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), where I work on cognitive cartography: multimodal representations that let people and AI reason over the same medical evidence with provenance intact.
Away from research, I teach as an Engineering Instructor at the Digital Career Institute, helping career-changers break into software engineering.
Contact
Prospective students, clinicians, and collaborators are welcome to write. Include a sentence on what you are working on; I read everything, even when the answer has to be no.
Curriculum vitae
My CV is available on request rather than posted here. Say who you are in the form above and I will reply with the current version.