Systems we build
Rillor builds and operates AI and data systems over large bodies of text and market data. Each runs in production and shapes how we build for clients.
Data
Market data warehouse
A research warehouse of U.S. equity and futures trade and quote history from 2018, alongside order-book data from on-chain perpetual futures venues, aligned on one clock.
Used to calibrate simulated markets, test forecasting methods and check rule models against recorded transactions.
- How it works
- Perpetual futures order books are rebuilt from snapshots and book diffs, with fills, liquidations and funding. Equity and futures trades and quotes share the same clock and symbol map, so events from different sources line up.
- For clients
- We take high-volume, time-sensitive data from capture to research-ready tables. Custom datasets and client research are built the same way.
Agentic AI
Grellum
Agentic legal research, independently branded and run at grellum.com. It answers research questions over a legal corpus, and every citation must match retrieved evidence.
- How it works
- In deep mode, a planner assigns parts of each question to four to six specialist agents, and an audit step reviews their work. Every citation must match retrieved evidence.
- For clients
- Multi-agent research over a large corpus can be held to its sources. We apply the same design to document, regulation and policy work for clients.
Infrastructure
The Rillor platform
The infrastructure every Rillor product runs on: Kubernetes, distributed storage, GitOps deployment and network-policy isolation between tenants.
- How it works
- Changes ship from Git. Each tenant runs in its own namespace, bounded by access control, admission rules and network policies.
- Isolation tests
- 109 access, admission and network tests passed for tenant isolation, run on 7 Oct 2026.
- For clients
- We build and operate private AI environments ourselves. Client deployments use the same patterns, on infrastructure the client controls.
AI tooling
MegaMCPs
Governed access to business tools for AI assistants, through an isolated MCP server for each customer.
- How it works
- Each customer's assistant connects to its own MCP server. Tool connections run inside that server, and no tool runtime is shared between customers.
- For clients
- AI assistants can use business tools with access governed per customer. Private AI cloud deployments include the same governed tool access.
Document AI
Public-records intelligence platform
A platform that archives public procurement records and regulations, extracts their document text and tracks how they change.
Public procurement recordsDocument and regulation intelligence
- How it works
- Records and their attached documents are archived as published. Text is extracted, with OCR for scanned pages, and each new version is compared with the last so changes are recorded.
- For clients
- Document AI at archive scale, from capture to extracted text to change history. The same approach fits any body of documents that changes over time.