By ALIS LABS
Valuation Copilot
Equity research and valuation on public data, with auditable models and an agentic copilot.
Context
Valuation is less math and more thesis: assumptions, narrative and consistency. Analysts lose time collecting public data, reconciling versions and rebuilding the same model.
Challenge
- Filings and earnings releases spread across different portals.
- Assumptions that are hard to version and defend.
- Natural-language questions that need data, documents and the web at once.
Solution
A data lake automatically fed with public CVM data and investor-relations documents, a valuation engine (DCF, dynamic WACC, Monte Carlo) and a copilot that decides when to query the database, the documents or the web.
Architecture
- 01Scheduled ingestion: public CVM data and IR documents
- 02Analytics engine: ratios, DCF, dynamic WACC and Monte Carlo simulation
- 03LangGraph supervisor with SQL, web search and RAG agents
- 04PostgreSQL, Qdrant for vectors, Redis for cache and rate limits
- 05Streaming answers (SSE)
Governance & security
- JWT authentication with roles (analyst, admin)
- Data validation at ingestion
- Copilot answers with the source consulted
Current state
- In evaluation
- Operating figures will be published once real data exists