Oct 2025 · Complete

InvestIQ — AI-Powered Investment Analysis

Gemini integration, prompt engineering, and frontend logic — team build

HackathonAI/ML

The problem

Built at HackNC 2025 for the Fidelity QuickVest challenge: help someone open their first investment account without drowning them in terminology. The hard part is not the account flow, it is that finance explains itself in language beginners cannot parse, and an LLM left unconstrained will happily do the same thing.

What I built

Integrated the Google Gemini API to generate investment insights inside the platform.
Prompt-engineered structured instructions controlling response format, tone, and analytical depth, which is what made outputs consistent enough to put in front of a user rather than merely plausible.
Constrained responses to a short length at a grade 6–8 reading level — a product decision enforced through prompting, on the view that an explanation a first-time investor cannot read is not an explanation.
Implemented the frontend logic that surfaces AI-generated insights in the interface.

What the project included

A Next.js and TypeScript application with a seven-step account-opening flow, built by the team.
Client-side document scanning, identity verification, compliance rules and audit logging.
Accessibility work including ARIA live regions, a high-contrast mode and adjustable text sizing.
A Playwright end-to-end test suite and a Prisma-backed data layer.

What I had to learn

Prompt engineering as interface design rather than model tuning. Format, tone and reading level are product constraints, and the prompt is where they get enforced.
Building against a sponsor brief under hackathon time pressure, where scope has to be cut against a deadline that does not move.
Working inside a larger codebase owned by several people at once, which is a different discipline from building something solo.

Stack

TypeScriptNext.jsReactGoogle Gemini APITailwind CSSZustand