Description of Organization
Tartigrade Ltd. (TTG) is a privately held firm specializing in options trading. We develop proprietary software tools that scan markets and trade automatically, with the goal of democratizing the finance industry by putting intuitive, powerful tools in the hands of individuals, small to mid-size hedge funds, and educational institutions at every skill level. Our long-term success is built on talented people who define our culture, live our values, and drive our purpose forward.
Problem area
Staff on TTG Talk, FEC's internal messaging platform, lose time and clarity when crafting responses during fast-moving discussions. Students will explore how conversational AI can reduce that friction without disrupting an existing, actively used communication tool.
Main objectives
Build an LLM-powered reply suggestion engine that reads recent conversation context and drafts tone-appropriate replies users can send, edit, or discard. Integrate suggestions inline with the native chat UI at an acceptable latency, and define a way to evaluate suggestion quality (such as acceptance or edit rate) so the team can validate real friction reduction. Stretch goal: extend into a voice-to-intent interface that transcribes speech into polished messages using the same context engine.
Scope of work
Review TTG Talk's existing architecture and messaging data model; design the context window and prompt strategy for the LLM suggestion engine; build and integrate the backend service; develop the front-end UI layer for surfacing suggestions inline; run usability testing with real or simulated conversations to measure acceptance and edit rates; iterate based on findings. Also an IP agreement must be signed giving FEC ownership of the product post-delivery.
Deliverables
- Report
- Presentation
- Website
- Resources
Frequency
- Bi-weekly
Skills and training required
- LLM/prompt engineering and API integration
- Full-stack web development (React/TypeScript front end, Express.js back end)
- UI/UX design for unobtrusive in-product features
- Basic experience with usability testing or evaluation methods
- Speech-to-text integration skills needed if pursuing the stretch goal
Resources required
- Access to TTG Talk's codebase and message data (within a defined context window)
- An LLM API or hosting environment
- The existing tech stack (React/TypeScript, Express.js, Keycloak, PostgreSQL)
- Staff availability for usability testing feedback
NDA or a commercialization agreement for this project?
Yes