/ij second meeting with Suranjit Adhikari from CapitalOne

Capital One AI Strategy and Role Discussion

Date/Time: March 16, 2026, 09:45 AM PDT

Overall Summary

The discussion covers Capital One’s AI strategy focusing on three key areas: AI evaluations, multi-agent conversational platforms (MCP), and advanced transformer-based customer modeling. The conversation also details organizational structure, decision-making, success metrics, tooling, and the interview process for an AI role.

Key Points
    •    Capital One is developing a unified AI evaluation platform for both Gen AI and traditional models to standardize and improve evaluation processes across teams.
    •    MCP governance is a priority to avoid tool overlap and confusion, with exploration of dynamic registry and code generation approaches.
    •    The company is building internal transformer models using extensive data to replace legacy models like XGBoost in credit markets, aiming for high impact with rigorous governance and audit.
    •    Capital One’s AI organization (IFX) operates horizontally, consolidating data science and engineering efforts with strong product team influence on budgeting and strategy.
    •    Success metrics include making Gen AI evaluations self-serve, deciding MCP rollout strategy, and expanding advanced model use cases while maintaining lean headcount.
    •    The interview process involves two technical rounds (system design focused on Gen AI and coding) plus behavioral and a unique case study assessing problem-solving and business acumen.

Action Items
    •    Candidate to prepare for interviews focusing on Gen AI system design, inference optimizations, and coding exercises.
    •    Vivek to coordinate with recruiter to schedule interviews and provide candidate support.

Open Questions
    •    Final decision pending on MCP strategy: whether to allow broad SDK use or limit to a few opinionated MCP servers.
    •    How to best balance tool design and governance to optimize agent workflows across diverse business units.
    •    Ongoing efforts to improve internal software delivery processes and reduce bottlenecks in deploying AI solutions.

---

/ij on the above, here are more detailed notes on the interview process and what to expect and how to prepare