Build and own an Investment Recommendation Specialist Model.
Should we buy, hold, or sell this position, and at what size?
Overview
Turns research and portfolio data into a specific buy/hold/sell action with a recommended position size instead of a raw research summary.
Who owns this decision
Chief Investment Officer, Portfolio Manager
Why owned AI may fit
Position sizing and buy/hold/sell calibration depend on a firm's own mandate, risk limits, and factor models, not general market knowledge a frontier model has.
What this Specialist Model actually does
Decision: Should we buy, hold, or sell this position, and at what size?
This is a larger system with 3 specialist model candidates and 1 supporting or external component.
Who uses this Specialist Model
- Chief Investment Officer
- Portfolio Manager
- Head of Research
- Director of Risk Management
Where the specialist model sits
This is a larger system with 3 specialist model candidates and 1 supporting or external component.
Quantitative model
Specialist Model
Quantitative model for expected return and position-size calibration using factor exposures and historical attribution
More Specialist Models in Financial services →
Long-document reasoning LLM
Specialist Model
Long-document reasoning LLM to extract structured signal from research notes
Specialist LLM
Specialist Model
Specialist LLM for rationale generation and investment-committee-ready summaries
Constraint layer enforcing portfolio mandate and risk limits
Supporting component
How Oumi connects to your stack →
How to evaluate and improve it
Evaluation
- Realized return of recommended positions vs. benchmark
- Recommendation acceptance rate by portfolio managers
- Calibration of expected return against realized return
Feedback signals
- Realized return of the position following the recommendation
- Portfolio manager acceptance, modification, or rejection of the recommendation
- Investment committee override and stated rationale
- Subsequent performance attribution vs. the recommendation's expected return
- Risk-limit breaches following a recommended position size
Model candidates to evaluate
- qwen3.5-9b
- glm-5.2
These are candidates, not performance claims. Run them against your evaluation set.


