Asymmetry Computing / founded 2025
We build PRISM: optimization infrastructure that connects numerical computation to the constraints of real institutional workflows.
ASYMMETRY COMPUTING / PRISMOur approach
A model is one part of a usable system. Inputs, deadlines, validation, and the record of a decision matter just as much.
Treat the incoming book, business constraints, and output contract as part of the product.
Evaluate timing alongside numerical quality and the conditions a decision must satisfy.
Publish scoped benchmark records and separate demonstrations from production commitments.
The company
Debdoot Ghosh
Founder, Asymmetry Computing
Asymmetry Computing was founded in 2025. Debdoot builds PRISM and is the author of Asymmetry PRISM: A CPU/GPU Portfolio Optimization Engine for Deadline-Bounded Institutional Rebalancing.
We are an early-stage company. Engagements begin with a defined workflow, representative inputs, and a test that makes the technical result clear.
Published work / official channels
The technical paper, published in June 2026, describes deadline-bounded institutional rebalancing.
The official repository complements the published research and benchmark record.
Company updates and the official identity behind PRISM.
PRISM on this website is the optimization infrastructure developed by Asymmetry Computing. It is distinct from unrelated products that use the same name. The research library and official channels above identify our published work.
Company & product
Reviewed
PRISM is portfolio optimization infrastructure developed by Asymmetry Computing, founded by Debdoot Ghosh. It supports direct indexing, tax-aware rebalancing, portfolio transitions, and constrained portfolio construction. The B2B2C model serves individual investors through advisors, registered investment advisers (RIAs), custodians, and wealth platforms.
PRISM addresses the construction and rebalancing decision: proposed portfolios subject to investment objectives and constraints. The job API describes how to submit inputs and retrieve results. Investment teams define their acceptance criteria and review outputs within their own implementation process.
The research library links the original PRISM paper and public evaluation materials. The benchmark record distinguishes generated numerical workloads, batched accounts, and executed-trade acceptance tests. Each result should be read with its workload, hardware, quality gates, and timing boundary.
Begin with a matched-workload pilot. Agree the data, constraints, comparison, and acceptance criteria before evaluating quality and runtime. Public benchmark results provide context; the pilot tests the fit for your specific workflow.
PRISM-Q is the quantum-simulation research workflow for discrete portfolio candidates, followed by classical refinement and verification. The published workflow does not establish physical QPU execution or measured incremental quantum advantage.
Work with Asymmetry