Trading Counterparty Recommendation System
A contextual recommendation system helping traders select counterparties and analyze transaction costs.
Business issue
Why I was brought into the project
Trading teams needed decision-support models to select counterparties, estimate pricing references and better understand execution costs.
Context
The environment around the project
Traders needed model-based support for counterparty choice and execution analysis.
Functional environment
The functional context covered counterparty selection, financial product pricing and transaction cost analysis for traders.
Technical environment
The technical environment combined Python, machine learning, Bayesian approaches, SQL, derivatives, fixed income and Bloomberg data.
Challenges
The recommendation had to account for execution context, market conditions, counterparty status and historical performance.
Solution
My contribution and its impact
My contribution to the project
I implemented a contextual Bayesian recommendation model, developed a regression pricing model and used both to support transaction cost analysis.
- Counterparty recommendation model
- Financial product pricing model
- Transaction cost analysis framework
- Trader-oriented analytical outputs
Impact
The work helped traders analyze execution choices more systematically and identify factors influencing transaction costs.
- More structured counterparty selection
- Reference pricing support for traders
- Clearer understanding of execution cost drivers
Impact metrics
Approach
How I structured the work
- Model the execution context and relevant counterparty features.
- Build a Bayesian recommendation model and pricing regression model.
- Analyze transaction costs to identify levers for future negotiation.
Takeaways
What I learned from this project
- Financial decision support models must reflect market context and user workflow.
- Recommendation, pricing and TCA become more powerful when designed together.
- Quant models need clear interpretation to be useful for trading teams.