ForSight.AI

Case summary page
StackAI workflow
Airia WorkFlow
Timeline - CaseBoard
Chat

Inspiration

In modern investigations, police and journalists deal with scattered case files, witness statements, and crime-scene summaries. Piecing them together into a coherent timeline is slow and error-prone. We wanted to build an AI copilot that could accelerate this process, turning raw reports into structured visual evidence and actionable insights.

What it does

ForSight.AI transforms unstructured forensic inputs, case files, testimonies, and crime-scene notes into AI-generated visual hypotheses, chronological timelines, and an interactive case graph. It also features a chat interface where investigators can ask questions, summarize findings, or even scrape related news using Apify, connecting real-world data with case reasoning.

How we built it

We used StackAI as the core workflow engine to orchestrate the entire forensic reconstruction pipeline. Every text input, whether it’s a crime scene summary, witness testimony, or case report, is passed through StackAI, which intelligently extracts key events, timestamps, and entities. Each event then generates a detailed caption that describes the visual scene, which is later converted into AI-generated imagery for visual hypothesis generation.

Airia powers the reasoning and chat layer. Whenever an investigator asks a question, Airia dynamically routes the query, deciding whether it should access internal evidence databases (sensitive files, event graphs) or external intelligence via APIs. This ensures context-aware responses while maintaining privacy and structured reasoning.

We integrated OpenAI models alongside our StackAI pipeline to refine language understanding, improve event coherence, and generate human-like narrative summaries that strengthen both the chat and visual reasoning experience.

Finally, our backend connects these components, event extraction, caption generation, image synthesis, and reasoning into a seamless loop that outputs structured graphs, timelines, and chat responses in a cinematic dark-themed UI that feels like a real forensic workstation.

Challenges we ran into

Designing a consistent event-to-image pipeline with high semantic accuracy. Managing multiple reasoning layers between StackAI and Airia. Handling sensitive data securely while enabling real-time retrieval. Aligning all outputs with ethical guardrails and clear “visual hypothesis” disclaimers.

Accomplishments that we're proud of

Built an end-to-end forensic reasoning system in under 5 hours. Created a timeline and caseboard graph that visually connects all events. Integrated real-time news scraping and a self-learning chat agent. Delivered a visually polished UI that feels like a real investigation dashboard.

What we learned

We learned how to combine multi-agent reasoning with visual generation pipelines, balance realism with responsibility, and design interfaces that maintain trust and interpretability in AI-generated evidence.

What's next for ForSight.AI

We plan to integrate live CCTV and weather metadata, add multi-case pattern comparison, and deploy ForSight.AI as a training and visualization tool for digital forensics teams. Future versions could also leverage voice and AR interfaces for immersive crime-scene reconstruction.