Research/Fig. 03 — two sources, one pattern
On the limits of what a machine may conclude.
First-author work on recovering production pipelines without a human in the loop.
Proposes a vendor-agnostic architecture for agentic incident triage, diagnosis, guarded remediation, verification and organisational learning across data and AI pipelines, assembled entirely from open-source components.
Downstream results produced by the autonomous research system I architected.
A double-blinded validation of the autonomous agent I architected, run against reference data to test whether the system's discovery process holds up when the answers are withheld.
Applies the tool-using agent and its distributed execution layer to broad-spectrum antibiotic discovery — a workload that depends on high-throughput candidate validation and ranking.
A safety evaluation of the agent's capability boundaries — what a capable tool-using scientific system can and should be prevented from doing.
End-to-end automation of antibody design and validation, demonstrating the agent running a full research workflow rather than a single assisted step.
Author positions are listed exactly as published. On the Moremi papers my contribution is the intelligent system and the distributed research infrastructure it runs on — the biology is the domain those systems were pointed at, not my claim to expertise.