Clarify

Inspiration

We were inspired by Obsidian’s graph-first way of thinking: ideas become nodes, and relationships become visible. We asked ourselves what that experience would look like for investigations, where evidence is scattered across many formats and important connections are easy to miss. Clarify came from that idea: combine a graph-native interface, AI assistance, and a strong ingestion pipeline so investigators can move from raw data to actionable insight much faster.

What it does

Clarify ingests diverse investigative artifacts, including emails, messages, metadata, reports, PDFs, images, and transaction files. The backend normalizes this data into a consistent structure, links related entities and events, and highlights potential contradictions or anomalies with evidence context. In the web app, investigators can explore the case as an interactive graph, drill into artifacts, and ask AI-guided questions to quickly understand relationships, test hypotheses, and clarify unclear threads.

How we built it

Frontend: Built with Next.js 16 + React 19 + TypeScript. Backend: Built with a FastAPI + Neo4j system. AI/ML: We integrated OpenAI with custom, investigation-specific orchestration so AI supports the workflow rather than replacing analyst judgment. Outputs are grounded in the underlying graph and evidence references.

Challenges we ran into

One of the biggest challenges was turning messy, inconsistent evidence (emails, documents, images, and transaction data) into a structure investigators could actually trust. We also had to balance speed with accuracy, making sure the system surfaced useful contradictions without overloading users with noise. Another challenge was designing an interface that makes complex relationships understandable at a glance.

Accomplishments that we're proud of

We built Clarify into a workflow that can quickly connect scattered artifacts and highlight high-value contradictions with evidence to back them up. We’re especially proud of creating a product that feels practical for real investigations, not just a technical demo. Most importantly, we focused on explainability so findings are transparent and defensible.

What we learned

We learned that in forensics, trust and clarity matter as much as model quality. Strong structure, clear evidence trails, and human-readable outputs make a much bigger impact than flashy automation alone. We also learned that users need guided triage first, then deeper analysis tools.

What's next for Clarify

Next, we want to improve signal quality with smarter prioritization and better investigator feedback loops. We plan to expand data-source support and collaboration features so teams can work cases together more effectively. We also want to harden Clarify for production deployments in enterprise and public-sector environments.