Forseen

Product Architecture

Inspiration

Regulatory compliance is one of the most complex, expensive, and opaque challenges businesses face today. Companies spend millions on legal counsel just to understand which regulations apply to them, and even then, they often miss emerging rules until it's too late. We saw an opportunity to democratize compliance intelligence, giving small and medium-sized businesses the ability to see what's coming before it hits. Forseen was born from the idea that AI can transform how businesses anticipate and respond to regulatory change.

What it does

Forseen is a compliance intelligence platform that transforms a company's profile into actionable regulatory insights. Users complete a guided onboarding wizard capturing their industry, data practices, operational footprint, and compliance posture. They then enter a compliance topic, like "health data privacy," and Forseen fetches live regulatory signals from sources including the Federal Register, Congress.gov, FTC, FDA, NIST, CISA, CMS, state legislatures, and news outlets. The platform decomposes the topic into specific sub-areas, generates risk predictions with probability and priority scores, and produces structured compliance reports with actionable recommendations. An interactive dashboard visualizes everything, and a RAG-powered chat interface lets users ask follow-up questions grounded in their analysis and the retrieved regulatory documents.

How we built it

The frontend is built with React 19, TypeScript, Vite, and Tailwind CSS, using Radix UI for accessible components and Framer Motion for animations. The backend runs on FastAPI with MongoDB for signal storage. We built two dedicated AI model servers: K2, powered by K2 Think V2, handles topic decomposition and regulatory predictions, while Hermes generates structured compliance reports and narratives. A data ingestion pipeline pulls from nine regulatory sources, Federal Register, Congress.gov, FTC, state legislatures, federal agencies (FDA, CMS, ONC, NIST, CISA), Reddit, Google News, and NewsAPI, computing signal velocity and vector embeddings for semantic search. The full stack runs locally via a coordinated script or through Docker Compose.

Challenges we ran into

Ingesting and normalizing data from nine wildly different regulatory sources, each with its own API format, rate limits, and data schema, was a significant engineering challenge. Building reliable topic decomposition that breaks broad compliance questions into meaningful, actionable sub-topics required extensive prompt engineering and iteration with the K2 model. Orchestrating parallel prediction generation across multiple sub-topics while keeping latency reasonable pushed us to optimize our async pipeline carefully. Ensuring the RAG chat interface could ground its responses in both the analysis context and the retrieved regulatory documents without hallucinating was another major hurdle.

Accomplishments that we're proud of

We built a fully functional end-to-end compliance intelligence pipeline, from raw regulatory data ingestion across nine federal and state sources all the way to actionable, company-specific predictions and reports. The platform handles real regulatory signals, not mock data. Our topic decomposition and prediction system reliably turns a broad compliance question into specific, scored risk assessments. The interactive dashboard presents complex regulatory information in a way that's genuinely usable, and the RAG chat gives users a natural way to explore their compliance landscape.

What we learned

We learned how fragmented and inconsistent the regulatory data landscape is, every agency publishes differently, and there's no universal standard. Building AI systems that reason reliably about legal and regulatory content requires careful grounding and retrieval, not just large language models. We gained deep experience with multi-service orchestration, async pipelines, and the nuances of building a full-stack product under hackathon time constraints. Most importantly, we learned that the compliance space is hungry for tooling that makes regulations accessible and actionable rather than intimidating.

What's next for Forseen

We plan to expand our regulatory source coverage to include international bodies like the EU, add real-time monitoring with alerts when new regulations relevant to a user's profile are published, and build out historical trend analysis so companies can see how regulatory pressure in their space is evolving over time. We also want to add collaborative features for compliance teams, integrate with existing GRC platforms, and fine-tune our models on domain-specific regulatory corpora to improve prediction accuracy. Long-term, Forseen aims to become the go-to intelligence layer that sits between raw regulatory output and business decision-making.