Principles

A short, honest statement of what we are, what we are not, and which customers we can serve well.

What we believe

Quant-native discipline — walk-forward, PBO, DSR, CPCV, bi-temporal data — is a platform property, not a configurable add-on. If it is optional, it will be skipped.
Infrastructure isolation beats application-layer multi-tenancy for production-grade workloads. Silo is expensive and correct.
Your backtest should survive its own promotion to production. _knowable_at filtering at query-build time is not discretion; it is a query-planner rule.
Research code and the production serving path must run the same function. Train-serve skew is a class of bug we close structurally.
The stack must be reproducible on a laptop. A cloud-only service that can't be emulated locally is disqualified from the architecture.
Open source is the default. Lock-in is easier to monetise but harder for a customer to trust; we chose trust.

Not a notebook-driven platform. Research code lives in a version-controlled Python package from day one, so promotion to production is the same code running in a new role — not a re-implementation of something a quant wrote in a cell.
Not a backtest library. vectorbt and QSTrader exist; we wrap them with PBO/DSR/walk-forward enforcement.
Not a research environment. Researchers do early-stage exploration anywhere; we accept research output, we don't dictate research process.
Not a workflow-DAG platform. Orchestration lives in-process — APScheduler for scheduling, PGMQ workers for queues, the same Postgres that holds audit and state. A DAG engine only enters the stack when a customer pipeline actually needs one.
Not a generic MLOps tool. Our defaults are quant, not medical imaging, not NLP for customer service.
Not a hedge fund. We don't invest; we don't take management fees; we don't have proprietary alpha. We are infrastructure.

A short list, said plainly:

Enterprise-tier hedge funds ($20B+ AUM, 50+ engineers). You can build this yourselves; you should.
Retail systematic traders. The floor cost of silo tenancy is wrong for you; QuantConnect and Alpaca serve that segment.
Multi-asset macro funds whose primary workflow is discretionary with systematic overlay. Our workflow shape is cross-sectional alpha; you would fight us.
Funds entirely on AWS who refuse GCP. GCP is our default substrate.

If you think we might fit, email us.