ML systems engineer · Independent researcher
I design production ML infrastructure and research causal information acquisition, replayable agent evaluation, and cost-sensitive control.
Selected research
Preprint · August 2026
When should an RL agent pay for more information? Causal Observability Optimization treats coarse observation, boundary observation, probing, transfer, continuation, and abstention as cost-sensitive meta-actions. The work includes a Lean-checked finite theorem, typed replay infrastructure, and bounded POSCM and MiniGrid validation.
Lean-checked · POSCM + MiniGrid · CC BY 4.0
Current research questions
01
When should an agent retain a coarse view, acquire richer context, actively probe, transfer, or abstain?
02
How can typed events, replay, and evidence manifests connect model behavior to scientific and operational claims?
03
How should long-horizon agents construct and navigate causal, temporal, and topic-indexed representations?
About
I am an ML systems engineer and independent researcher working between production infrastructure and experimental AI. Over the past decade, I have built enterprise software, MLOps platforms, applied machine-learning systems, and multimodal models, while leading technical work across AI and ML programs.
My current research focuses on causal information acquisition, replayable agent systems, and evidence-backed evaluation. I am particularly interested in systems that connect formal claims, experimental execution, and reliable deployment.
Connect
For research, collaboration, or professional inquiries, email me directly.
contactme@kalielwilliamson.com