Retail investors want answers about listed companies in plain, understandable language. Companies want to gain insights about their investors. A newly founded startup wanted to build a platform for both, providing grounded AI answers in real-time to investors and uncovering investor opinion for companies. Modulai delivered the entire GenAI pipeline and cloud backend for the investor chat in 6 weeks.
Outcome
6
Weeks from idea to production
1000+
Listed companies supported
Challenge
Swedish retail investors struggle to interpret financial disclosures, annual reports, and press releases. Listed companies, on the other hand, lack visibility into what their investors are actually asking and how engagement varies with news and reports. The startup set out to solve both problems with a single platform. The challenge was trust: an investor relations product that gives unreliable answers doesn't just fail, it does active harm. Answers had to be fully grounded in approved public sources across 1,000+ listed companies, regardless of their IR budget or how much information is publicly available about them. At the same time, the startup needed a fast launch in order to start testing their product on the market.
Solution
Modulai designed and built the complete GenAI pipeline and cloud backend, with a streaming API, LLM integration, GCP infrastructure, and CI/CD. The chat streams answers in real-time, discloses its sources for transparency, and prompts users with relevant follow-up questions. It’s a custom chat with a clear purpose, and thus refrains from giving investment advice or answering questions unrelated to listed companies and investments. A tailored system prompt is constructed per company, so the agent functions as a knowledgeable expert for whichever company the user asks about. The architecture is provider-agnostic, but currently running with Gemini on GCP Vertex AI.
Tools
The service is built on Python and FastAPI, streaming responses via Server-Sent Events. The production LLM is Gemini on Vertex AI, using Google Search and URL Context tools to ground answers in real time. The service runs on GCP Cloud Run, deployed through GitHub Actions. Observability is handled via Cloud Trace with OpenTelemetry GenAI semantic conventions, giving full per-request audit logs. Output quality is tracked with an offline RAGAS evaluation pipeline, with results logged to Vertex AI Experiments.
Value created
For retail investors, the platform removes a real barrier: complex financial information is now accessible in plain language, on demand. Having this live enabled the startup to start collecting valuable data on which questions come up, how sentiment shifts around earnings and announcements, and where communication gaps exist. Modulai delivered the full AI component in six weeks, enabling the startup to go to market faster than a custom build would typically allow. The provider-agnostic design means the startup isn't locked in to a single model vendor, rather allowing costs and capabilities to be optimized as the market evolves. Thanks to the rapid delivery of a production-ready system, the startup hit the market fast and is now learning from real users, the only way to truly find product-market fit.
Learn more
