Autodidact
An AI tutor for studying on your own.
I lead the design and engineering of Autodidact. Learners work through reading and practice with Faraday, then return to what needs review. Saved Python exercises give the tutor their work to discuss.
MARCUS WILLIAMS · AI SYSTEMS & APPLIED AI
I build AI systems
that do useful work.
I build agents, train and evaluate models, and turn ideas into working software. My work spans independent products, research engineering at Intel, and open-source systems built with teams across companies.
01 / INDEPENDENT PROJECTS
Ideas, put to work.
Syndicate
An operating system for computational organizations.
I’m building Syndicate so people can create institutions whose everyday work is done by agents. Its kernel tracks authority, commitments, completed work, and whether the work achieved its purpose. The system is pre-release.
Evolved
An AI-assisted algorithm discovery platform for turning code search into a repeatable, auditable experiment.
I designed and built the search and evaluation system. It generates candidate programs, tests them in isolated workers, and keeps their results and ancestry. Separate checks have exposed candidates that improved the score by violating the intended arithmetic.
Sitrep
Team reports that keep dependencies visible.
I’m building Sitrep to collect progress, blockers, and next steps through the web and email. It generates project, group, and organization summaries with the original updates available to read. Evaluations check whether important facts survive those summaries.
02 / MODEL WORK · INTEL NCL
Training and evaluating
sparse language models.
At Intel’s Neuromorphic Computing Lab, I pretrained and instruction-tuned 340M- and 1.3B-parameter HGRN language models at 90% sparsity. I evaluated checkpoints with custom tools and standard benchmarks as part of the lab’s neuromorphic model work.
Model work and professional experience ↗- 340M & 1.3B
- Model parameters
- 90%
- Sparsity
- Train. Evaluate.
- Pretraining, instruction tuning, checkpoint evaluation
03 / ENGINEERING & DELIVERY
Built with teams. Used by teams.
Software people use.
At Intel NCL, I built a DevKit adopted by the lab that reduced the path to a first workload from months to about 30 minutes. I also built Lava’s CI, hardware tests, and release pipelines into the lab’s regular workflow.
Delivery across teams.
I led two engineers and coordinated architecture and integration work for ONAP releases across projects and companies. I contributed as a committer across four projects and was selected by the TSC as a Super Committer.
Performance, measured.
My work includes upstream OpenJDK changes and tests of 1,800 OVSDB nodes against one OpenDaylight controller, covering registration and teardown.
THE PERSON BEHIND THE WORK
A generalist by nature.
A builder by practice.
My work has taken me from open-source networks and Java performance to neuromorphic computing and AI research. The through line is the same: understand the system, build something, and test what you think you know.
A little about me →COMPARE NOTES
What are you working on?
If you’re building something useful with AI, I’d like to hear what you’re working on.