AI Finance

We minimised burn rate through bootstrapping — small team, tight scope. I handled design and product while my co-founder managed engineering and business. No fluff, no runway to waste.

Research started with interviews with hedge fund analysts and office visits, plus tool audits of Bloomberg and Koyfin to understand what serious financial professionals actually needed from a research tool.

We built an AI-first web app for stock research — replicating the interactive depth of a conversation with a human analyst. The core was a knowledge graph synthesising both qualitative and quantitative data, giving users a way to explore companies that felt more like a dialogue than a dashboard.

GPT-3.5 hallucinations proved a real problem early on. Reliability was non-negotiable in finance, and established giants with massive data moats made the original direction increasingly difficult to sustain.

We identified a gap: IR websites overwhelm investors with dense information. The pivot was a chat-based solution for investor relations teams — Ralina, an AI persona that could be embedded on IR sites, with voice interaction and avatar customisation.

IR teams had minimal budgets and were highly price-sensitive. SEC compliance added more complexity than we could handle at our scale. So we pivoted again — this time to KV-01, a sentiment-driven investing platform that personalised companies as fruits, flowers, and spices to engage beginner investors emotionally rather than analytically.

We hit 20–30 MAUs in the first month, but recognised the sustainability challenges early. The biggest lesson: we over-indexed on fundraising conversations when we should have been shipping and testing faster.

Navigating real-world constraints around data, distribution, and direction across multiple pivots taught me more about product instinct than any single-direction project could. Startup design isn't just about craft — it's about knowing when to kill an idea and when to back it harder.