Formula
Pricing notes, risk formulas, Greeks, and curve logic become typed Chelis programs with source traceability.
See the pathHigh-stakes numerical systems
Verified numerical code
AI agents are producing more numerical code than review can absorb. C Proof moves consequential kernels from Python's runtime boundary into Chelis, a checked substrate where shape, precision, effect, and ownership rules are enforced before production. Property and proof checks bind implementations to stated specifications, with evidence your team can inspect.
Why now
Agent-written numerics
A model can draft a day's worth of pricing, signal, risk, or data-transform code before a human can read one kernel closely. Python accepts too much of that ambiguity at run time, after the code has already crossed into the research or production path.
C Proof gives generated code a checked substrate. Move the consequential kernels into Chelis, keep typed constraints and specification properties in the build, and leave lower-risk Python around them while migration happens function by function.
Inputs
Checked path
Python
Existing numerical functions can be translated function by function instead of replacing the research stack at once.
See the pathSpecification
Portfolio constraints, risk measures, and signal rules become explicit program properties.
See the pathTyped tensors
Explicit tensor types and pipe stages keep the numerical path visible to the type checker.
def pipeline(x: &tensor[n, f32]) -> tensor[n, f32] =
x |> relu |> sigmoidChecks
Trust stack
Type rules
Chelis type rules reject shape, precision, effect, and ownership failures before a binary is produced.
Properties
Program properties bind the implementation to a stated specification. SMT discharges solver-decidable obligations.
Artifacts
Complex properties stay visible as property-based validation, with preconditions, sample count, and provenance recorded for review.
Numerical libraries
Chelis-native coverage
Chelis-native libraries cover the numerical work teams usually reach for in Python, C++, and notebook ecosystems, while keeping critical constraints inside the checked program.
- Greeks and pricing kernels
- Yield curves and stochastic processes
- Risk measures and portfolio constraints
- Signals, optimization, and linear algebra
Deployment
Controlled boundary
C Proof is built for teams that cannot send proprietary models, positions, or build paths through an external service.
- Runs on infrastructure you control
- No hosted dependency for proprietary models or positions
- Compiler source visible under enterprise agreement
- No public registry in the build path
Teams
Accountable owners
Quant research and trading leads
Move generated pricing, signal, and risk code through checks before it reaches production.
Engineering leadership
Give AI-generated numerical code a narrower path into controlled systems.
Model risk and governance
Review specifications, proof artifacts, provenance, and deployment records in one chain.
Security and compliance
Keep source visibility, supply chain control, and auditability in the deployment plan.
Next step
Contact
Bring a pricing kernel, risk measure, signal path, or generated numerical function. We will map the checks, artifacts, and deployment boundary it needs.