Not just models — the framework, the schedulers, the registries, and the dashboards that turn a quant's hypothesis into a monitored production strategy. This portfolio doubles as the operator workspace:/desk is the morning briefing, /strategiesis the live catalog, /featuresis the leaderboard. Drive your day from the site.
$ rb run alpha.xs_momentum --universe synth ▸ resolving features ok ▸ training ok walk-forward · 5 folds ▸ evaluation ok rank_ic 0.0297 ▸ registry ok model xsmom@sample-r1 ▸ promotion held awaiting named approver $ ▌
Morning briefing, strategy catalog, feature leaderboard, thesis-lifecycle view, endpoint reference — researchbook as the research engineer’s day-one surface (ADR-0059).
Open the desk →Ingest, features, training, registry, deploy, monitor — with the contract at each seam.
See the system →Feature graph, model registry, deploy contract, notebooks-as-code, alpha-bench — design decisions written out.
Read projects →Cross-sectional momentum, L1/L2 microstructure — fit in 600 ticks, update on slider drags, leakage toggle exposed.
Open sandbox →Eleven MCP tools, a published refusal vocabulary, and a runnable recovery example. The planner branches on typed slugs, not prose.
See the posture →Research platform & alpha-model framework
Feature engineering & data pipelines
Training, registry, deployment, monitoring
Cloud, containers, observability, CI/CD
Quant + trader feedback loops
The framework you spend most of your day in. Declarative model definitions, walk-forward CV, leakage-safe feature evaluation, and a benchmark grid that lets a quant compare three weeks of model variants over coffee.
Treat features like a build system: declarative dependencies, point-in-time correctness baked in, content-addressed caching, and the same graph at training time and serving time.
The production half of the loop: the same feature graph that ran in research, now consuming a market-data stream and emitting signals into the trading stack — with point-in-time guarantees, lineage, and a kill switch.
A self-hosted model registry that's the source of truth for what trains, what ships, and what's running. Each model carries its full lineage — code commit, feature graph hash, data window, owner, evaluation report — and the production platform refuses to load anything it didn't register.
A small library and a sidecar that let a researcher prototype freely in a notebook, then promote the cell that worked into a versioned, reproducible pipeline — without rewriting it. The notebook is the IDE; the platform is the runtime.