Managed compute for research desks
Everything in Pro for 5 seats, plus a managed compute plane: cloud sweeps, Monte Carlo and walk-forward on our infrastructure, hosted data credits, and a direct support channel. Credits included, overage capped by default.
How a job runs
Interactive work stays local and instant. The heavy work is submitted, routed, metered, and comes back small.
- 01your machineThe sweep leaves the notebook, the notebook stays freebt.cloud.submit(sweep)
- 02managed capacityRouted by size: warm CPU for interactive runs, scale-to-zero GPU for big sweeps, Monte Carlo and walk-forwardCPU / GPU
- 03back to youResults land in the session that submitted themmetrics + top-N
Every job writes one metered line to your ledger: what ran, for how long, for how many credits. The exact SDK surface is being finalized with the first early-access desks.
Compute that is not your laptop
The workloads that decide whether a strategy is real, full sweeps, deep Monte Carlo, walk-forward across the whole book, are exactly the ones a laptop is worst at.
Cloud parameter sweeps
Sweeps that would take hours on a laptop run on managed GPU capacity. Submit from the SDK, keep working, get the results back.
Monte Carlo at scale
Bootstrap and permutation studies with millions of resamples. Tail risk and P(ruin) estimates without tying up your own machine.
Walk-forward batches
Run walk-forward across every strategy you have, in parallel, and compare degradation curves side-by-side.
Screen fast, verify exact
A fast screening pass ranks the parameter space, then the top candidates are re-verified on the exact engine path automatically.
Credits, with a hard ceiling
The plan includes a monthly allowance of compute and data credits. Everything is metered per job, visible in real time, and capped: by default, spending beyond the allowance is simply blocked.
How credits work →Compute and data each come with 5,000 credits a month, and both follow the same path: an alert fires at 80%, and past the allowance, jobs are refused rather than billed. The pay-as-you-go zone only exists if you opt in, under a hard cap you set yourself.
Spend caps by default
Overage is blocked until you opt in to pay-as-you-go, and a configurable hard cap bounds monthly spend. No surprise invoices.
Usage dashboard
Balances, history per job (what ran, when, for how many credits), end-of-month projection, and your API keys, all in one page.
Data credits, cached
Query Databento and Massive through your data credits. Data already served to your team is cached and not billed twice.
Metered per second, in credits
Compute is billed per second of machine time by machine class, data per query, both in credits. The rate card is being set with the first desks and goes up on this page before anyone is charged for overage: priced from a list, not from a call.
Five seats, one bill
Every researcher on the desk gets a full Pro license: 1-minute output, unlimited sweeps and Monte Carlo, WFO, safety checks, GPU acceleration, Databento and Massive connectors. Need more seats? They are extendable, tell us how many.
Beyond Team
SSO and audit logs, on-prem or VPC deployment, custom integrations, negotiated data volumes and contractual SLAs live in the Firm tier, priced to your desk.
What Firm adds →FAQ
Get in early
We are onboarding a small number of teams and shaping the compute plane with them. Tell us about your desk and your workloads, and we will set you up.