Preserves source truth, standardizes signal structure, and shapes stable analytical outputs through bronze, silver, and gold layers.
Burns & McDonnell Interview
SignalOps Demo Portal
Built after comparing the role stack to my direct experience and choosing one project that would force me to work through layered data shaping, experiment tracking, serving contracts, and frontend delivery in one place.
What you can review here
Masked demo data, working MLflow evidence, stable frontend contracts
Public identifiers and sensitive details are masked, but the architecture, experiment workflow, and UI integration are real and directly inspectable.
Rather than describe the stack abstractly, I built a working artifact that shows how those concerns fit together in a coherent product.
Source
Bronze
Silver
Gold
Experiments
Serving
Frontend
Compresses a large signal set into a smaller explainable shortlist with priority score, tiering, and review context.
Uses local MLflow runs, metrics, and best-run summaries so the project has a real experiment history rather than a single notebook result.
Narrows the public surface to stable summary, detail, replay, live-context, and model snapshot contracts so the UI stays clean while research continues underneath.
Presents the product as a usable React interface with landing, demo, architecture, model, and walkthrough routes instead of requiring the reviewer to imagine the product shape.
Tracks structured workflows, waits, and checkpoints so multi-step work stays durable and reviewable instead of collapsing back into ad hoc notes.
The implementation is local-first, but it was intentionally designed to map to enterprise data platform, experiment lifecycle, and business-facing application concerns.
The project does not claim automated trading execution or enterprise ownership of every platform service. It does claim a real layered system, a real experiment loop, and a real product interface built for review.