Agentic research systems
Built from zero for investment research: architecture, agent harness, sandbox, orchestration, memory, tool use (MCP) and multi-agent workflows.
AI lab for finance · Research, advisory and build
Know what AI can be trusted with.
We design, build and test agentic AI for hedge funds, banks and asset managers. Evidence, not assurances.
What we build
Designed, built and shipped hands-on, from architecture to working system.
Agent-first engineering
Development teams reorganised so agents write the code and people own acceptance, including rollout of tools such as Claude Code and Codex.
Research knowledge base
Ingestion, structuring, retrieval and continuous updates, so agents work from reliable knowledge and context.
Evals and model selection
Benchmarks and evaluation datasets built on your own research tasks, used to choose models and tune prompts, context and harness.
Who we work with
Hedge funds and asset managers
Portfolio managers, analysts and the engineering teams behind them.
Banks
Risk, pricing and case-management workflows that can only change one step at a time.
How we start
Agentic Transition Assessment
For workflows too complex to replace in one go. Fixed scope, four weeks.
01
Map the workflow
Every step today: what systems such as recommenders do, what people do, and where they hand over.
02
Find the first step
Which steps can be checked against past outcomes, so an agent’s work can be verified before it is trusted.
03
Agent trial
Frontier agents run that step on historical cases. We measure where they match past outcomes.
04
Sequenced roadmap
Which steps change first, which stay with people, and the checks each change needs before it goes live.
Products
In private beta
Evaluation environments
Private, deterministically graded environments for teams training and testing agents on financial tasks. A private task library, built on financial code no model has seen.
Join the beta listFounder
Dr Alicia Vidler
Founder, VaLion Labs · Hong Kong
25 years where quantitative finance meets AI. She has validated the models banks rely on, built AI systems that traded real capital, and published peer-reviewed research on AI agents in financial markets.
Builds agent systems hands-on
Designs, builds and trains agentic workflows from scratch. Former Global Head of AI at Capitolis.
Agentic AI research
PhD in agentic AI (UNSW). Published at AAMAS, AAAI and IEEE. Author of a Springer Nature book on AI for bond markets (2026).
Knows the buy side
Co-founder and CIO of an AI-driven hedge fund. Senior Director, Global Prop Trading at Merrill Lynch / BofA. Built risk models at RiskMetrics for 200+ hedge funds.
Contact
Start with a conversation.