AI Enablement Engineer, Chicago area
I make software work inside real organizations.
I build AI systems that go into real use, then train the people who run them. It comes from ten years in and around local government, across 100+ court and payments rollouts.
Three things I built
All work
CourtCollectClient workLive demoA Texas municipal court's review console.

A person clears each flagged case before the collections agency sees it.
A review console a Texas municipal court's reviewer can finish alone: a person clears each flagged case before the collections agency sees it. Built around the reviewer's own checklist, with legacy court data mapped in and a PII boundary the tests enforce.
- Film
- 0:26, narrated
- Receipts
- 7 pull requests cited
HeddleProduct100% pass rateA certifying body's blueprint becomes an item bank, a weekly curriculum, and student-ready documents.

A 100% pass rate across three cohorts.
Turns a certifying body's blueprint into an item bank, a weekly curriculum, and student-ready documents an instructor can teach from, with an assistant that shows what every action will do before it runs. In use for a community college exam-prep program with a 100% pass rate across three cohorts.
- Film
- 0:26, narrated
- Receipts
- 8 pull requests cited
Learn UkrainianCourse12,000 words studiedA free course from Pre-A1 to C1.

Built on an open frequency study of 12,000 Ukrainian words.
A free course from Pre-A1 to C1, with 44 units and 326 lessons and spaced review that saves progress on the learner's own device. It is built on an open frequency study of 12,000 Ukrainian words.
- Film
- 0:26, narrated
- Receipts
- 10 pull requests cited
Each build has a short film, narrated by an AI voice. AI voice and music by ElevenLabs. Photographic scenes are AI-generated. Stamped page shots are recorded from jakelawrence.xyz. The CourtCollect short is recorded from demo.courtcollect.xyz on invented data; delivery there is simulated.
AI systems, running in public
Each one is open to try. The scores re-derive in your browser and the answers cite their sources.
- EvalsBenchmarksReproducible benchmarks for LLMs and retrieval, this site's own AI features included. Deterministic scores re-derive live in your browser.Open
- Agent memorySAGEN-Bench v2A pre-registered test of an AI agent memory architecture on 72 frozen scenarios. The memory works. Perception is the bottleneck.Open
- Cited retrievalRules StackEvery rule stacked on a home in one Illinois village, each linked to its source, with an agent that answers from the corpus and cites it.Open
- Model gatewayInferenceThe models this site routes to, with list prices, context windows and probe-measured speed, behind an OpenAI-compatible gateway.Open
Games
All games34 interactive games: 9 teaching graduate-level mathematics, 4 exploring core AI concepts, 19 standalone skill games, and 2 language learning games.
Transformer ArchitectureThe Word MachineBuild a language model from nothing: tokens, attention, layers, prediction.Play
Goodhart's Law & RLHFThe Rating GameRate eight pairs of AI answers, watch your picks fit a reward model, then turn up the optimizer and see it learn your tells instead of quality.Play
Chess TrainingChessGuessrWatch an opening unfold and name it the move it splits from the rest, or take the name and play it.Play
Unhinged ArcadeFLAPPY-ISHThe one-button bird game with the safety off.Play
Selected writing
All writing- How I'd Roll Out AIGuideOctober 2026
- AI Acceptable-Use PolicyTemplateSeptember 2026
- SOC 2 Readiness for a One-Person PlatformAssessmentSeptember 2026
- Case Study LibraryDataSeptember 2026
- Seeing Like an AI CompanyEssayApril 2026