We built a DEX aggregator on Base, a consumer fintech app, a HIPAA document pipeline, an IL4+ GovCloud deployment and a fleet maintenance platform. Each case links the client's live site.
50+
Products shipped
40+
Clients
85%
Repeat clients
6
Audits published
The problem
O1 Exchange needed a swap router on Base. It had to find the best price across 15+ venues and complete each swap in milliseconds.
How it works: O1 Exchange
An order goes into the quote API. Beam search drops weak routes. The convex optimizer splits the order across venues: two shown, plus 13 or more. The Solidity router batches the legs into one swap.
Beam search drops a weak route before the optimizer splits the order.
What we built
Multi-hop routing with permit-enabled swaps.
The o1 launchpad on Uniswap v4, with locked liquidity and RWA tokenization.
The result
On benchmark trades, the router's prices stay within 30bp of the top aggregators.
DreamOS started from zero. Real money had to move through several providers without a dropped transaction.
How it works: DreamOS
Money makes a round trip. Fiat comes in through an on-ramp provider and becomes digital assets. It goes back out through an off-ramp provider as fiat. Each provider step has explicit error states. Options trading branches off the digital assets.
A provider step fails. The app holds it in an explicit error state, and no transaction drops.
What we built
The fiat to digital asset to fiat flow, with explicit error states at each provider.
Options trading in the same app, on React Native and Expo.
The result
The app handled $5M in notional volume on launch day, with zero dropped transactions.
The pipeline parses each patient document and extracts its fields. Each field gets a confidence score. High scores go on to validation. Low scores go to a person first, then to validation. Validated fields reach the patient record. Audit trails cover the pipeline.
A low-confidence field goes to a person before it reaches the record.
The problem
Hike Medical needed AI to read unstructured patient documents. National buyers audit every part of a system like that before they sign.
What we built
Fallbacks and live monitoring across the pipeline.
Audit trails, a HIPAA security posture and patient status dashboards.
The result
Extraction accuracy holds at 95%. Hike Medical then won Hanger, a national clinic chain.
ProDex Labs builds a systems engineering platform. Each defense contractor that joined it needed a one-off GovCloud deployment.
How it works: ProDex Labs
Each new contractor enters one automated onboarding path. The path crosses into the GovCloud boundary, which meets IL4+ and holds IAM policies, network topology, audit logs and encryption. The contractor comes out running on GovCloud.
Before: each defense contractor needed its own one-off GovCloud deployment.
What we built
One GovCloud architecture, made for IL4+ from the start.
Architecture leadership and hands-on training for the ProDex engineers.
The result
One repeatable path replaced the one-off deployments. It meets IL4+, the DoD level for controlled unclassified information.
Carma is a YC-backed fleet maintenance company. Fleet managers needed one place to request service, compare offers and track each vehicle.
How it works: Carma
Five telematics feeds, Geotab, Samsara, Azuga, Linxup and Raven, pass a VIN check that also writes an audit row. One vehicle record comes out. In the service loop, service centers bid on each request. An AI check reads the invoice. A manager approves the spend, and Stripe pays the work order.
A bad feed fails the VIN check. Nothing reaches the vehicle record.
What we built
Carma V2 on 4 surfaces: backend, fleet web, admin web and mobile webviews.
ServiceTitan, TireConnect, Motor and Recall Masters feed the same work order.
Email ingestion for the shops that still reply by email.
The result
A bad feed cannot corrupt a fleet's history. We also helped Carma get SOC 2 Type II.
“XORS was great to work with. They took our input and strived to create a great product. Great at communicating and responding to tight deadlines, with real attention to detail.”
“They anticipated needs we hadn't even voiced. We knew it was working once we were live processing orders, and now we have real process automation across several business functions.”
AnodyneBobby
“We've had an excellent experience with the XORS team. They provide valuable insight and proactively look for ways to help us meet our goals.”
Zero Cool LabsRichie Humphrey (devtooligan)
GoogleInternal data-processing toolingAI · 2024
State StreetInstitutional trading toolingFintech · 2026
