Work Term 3

INTRODUCTION

Throughout the Winter 2026 semester, I had the amazing opportunity of interning at Picton Investments for my third co-op work term as a Software Developer. This term, my work focused on building an MCP server for one of our applications so that we could expose our API to MCP clients like Claude, Codex, and Cursor. Most of my development work used TypeScript, with React for frontend work, SQL where the project touched data, and Azure DevOps for source control and team workflows. The goal was to let users access our features and product knowledge through the workflows and AI tools they already prefer.

EMPLOYER

Picton Investments is an investment management firm where software supports teams working across portfolio, risk, data, operations, and client-facing workflows. Specifically, I interned as a Software Developer and worked on tools that helped expose product capabilities through modern AI-assisted workflows. The company also had a strong intern culture, including the Picton NextGen event, which gave interns the opportunity to connect with each other and with staff across the organization.

GOALS

My goal for this work term was to grow from building strong web application features into designing software that could be used reliably by AI agents and LLM-powered clients. I wanted to understand MCP concepts deeply, learn how to model an existing API through MCP primitives, and build something that preserved the usefulness and intent of our product while fitting naturally into AI-assisted workflows.

JOB DESCRIPTION

As a member of the Software Development team, my responsibilities included:

• Building an MCP server that exposed application API capabilities to MCP clients.
• Designing TypeScript interfaces and tool behavior that represented the underlying API clearly and reliably.
• Testing MCP behavior thoroughly to ensure consistent responses and faithful handling of product features.
• Keeping token efficiency in mind when shaping responses for LLM-powered workflows.
• Using Azure DevOps for team development workflows and working with SQL where data access was involved.
• Helping with frontend design flaw backlog items in React in collaboration with the UX intern and team members.

CONCLUSION

During my third work term at Picton Investments, I had the opportunity to work on a project that sat at the intersection of APIs, AI tools, and product design. Building an MCP server primarily with TypeScript taught me how important it is to design clear primitives, expose capabilities in a way that is easy for clients to reason about, and keep token efficiency in mind when building AI applications or MCP servers. I learned that even small response design decisions can affect how useful, fast, and cost-effective an AI workflow feels to the end user.

The project also reinforced the importance of thorough testing. Since the MCP server needed to faithfully capture the essence of our API, I had to think carefully about consistent behavior, reliable outputs, and edge cases that could confuse a client or an agent. Outside of my main project, I also offered to help with our frontend design flaws backlog in my spare time, using React and working closely with the UX intern and other members of the team to make sure improvements were thoughtful and practical.

This term also gave me a much broader view of Picton Investments as a business. I had discussions with interns and full-time employees across many departments, learned more about areas like funds, securities, holdings, portfolios, issuers, exposures, risk drivers, benchmarks, and compliance workflows, and had the chance to speak with the AI team about MCP architecture, LLMs, skills, agents, and related ideas. Overall, this work term helped me connect implementation details with product value, user workflows, and the larger business context behind the software I was building.

ACKNOWLEDGMENTS

I am very thankful to everyone at Picton Investments who made this work term such a valuable experience. Thank you to my manager, mentor, the engineering team, the UX intern, the AI team, the other interns, and the full-time employees across departments who took the time to answer questions, share context, review work, and help me understand both the technical and business sides of what we were building.