Solomon Eshun·Founder, Lumis
Building intelligent systems that can explain their own failures — production MLOps, evidence-grounded agents, financial ML and applied research.
$ entropy always wins — the trick is choosing where it lands
Making computers go beep boop—reliably, at scale, and under real-world constraints.
EnBW Energie Baden-Württemberg AG (via Ishango.ai)·Data Platform Engineer → MLOps Engineer
Keeping production ML and data workflows alive across multiple teams at a large European energy utility — deployment, observability, and working out why a pipeline died at 3am.
First author
A vendor-agnostic architecture for agentic incident triage, diagnosis and guarded remediation across data and AI pipelines — assembled from open-source parts rather than bought.
Lumis·Founder
Finds why a data or ML system broke — from its own alert to the change that caused it — and tests every explanation against evidence it gathered read-only. Across 117 investigations with four models on the GridCast estate, 11 of 15 injected faults were never missed.
Noeud·Founding Machine Learning Engineer (R&D)
A probabilistic FX intelligence backend for Ghanaian importers and exporters: 30-day forecasts with block-bootstrap uncertainty, and an LLM layer that weighs them against market news and bank rates while model baselines stay immutable and audited.
MinoHealth AI Labs·Machine Learning Engineer
An autonomous tool-using research agent and the distributed platform under it — 20+ scientific tools behind one API, multi-day runs, 20,000+ candidates per batch. It produced four publications.
Evidence, uncertainty, and the boundary between machine inference and human judgement.
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.
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.
Talk is cheap. Here is what made it into production.
Building the intelligence layer for complex systems, starting with data and AI: an investigator that finds why a system broke and shows the evidence.
Sitting between data scientists and production: moving completed ML work into operable systems, and keeping them running.
Building the ML and intelligence layer for FX-risk decision support for African businesses.
Notes from the lab: methods, results, and useful failures.
117 investigations of injected failures, four AI models, three days — the results, the wrong answers, the bugs that were ours, and the limits of what it shows.
Read ↗The first version, part by part: what it does today, what it deliberately doesn't do, and what comes next.
Read ↗A realistic forecasting estate, broken fifteen ways, and six investigators — from a bare language model to the full Lumis loop — asked to find each cause.
Read ↗
I am mostly interested in what happens at the edge of what we can currently explain.
Systems fail in ways nobody designed for, and that failure is usually the most informative thing in the room. I build intelligent systems that learn from that gap, and I am heading toward cyber-physical systems: grids, factories and vehicles that must understand their own state and recover safely.
Curious about reliability, uncertainty, and how much a machine should be trusted to decide on its own. Open to research collaborations and hard problems in self-adaptive and cyber-physical systems.
