Case Cabinet

What it does

Case Cabinet is an AI-powered evidence organization and cross-referencing platform. Users can upload different types of evidence, including documents, images, audio recordings, and structured data, and organize them into a centralized case workspace.

Specialized AI agents process different evidence types, extract relevant information, classify content, and identify potential relationships between pieces of evidence. Case Cabinet can also perform public web research using search APIs while maintaining source references for retrieved information.

Our goal is to transform scattered files and information into a structured, searchable case workspace that helps investigators identify relevant details and connections while preserving the original evidence.

How we built it

We built Case Cabinet as a full-stack application with a modular backend architecture. The backend manages evidence uploads, file processing, classification, AI agents, cross-referencing, and external web research.

Different AI agents handle specialized tasks, including document analysis, image analysis, audio transcription, evidence classification, web research, and cross-referencing. Each component has a defined responsibility, allowing the system to process different evidence types through specialized workflows.

We also integrated SerpAPI for public web information retrieval and designed the architecture to support different LLM providers. Original evidence is preserved separately from extracted information and AI-generated results.

Challenges we ran into

One of our biggest challenges was figuring out how to process different types of evidence within a single system. Documents, photographs, structured datasets, and audio recordings all require different processing pipelines, but their extracted information needs to come together in one case workspace.

Another challenge was designing a multi-agent architecture where every agent has a clearly defined responsibility rather than relying on one AI model to handle everything. We also had to consider source tracking, uncertainty, and how to distinguish original evidence from AI-generated information.

Accomplishments that we're proud of

We are especially proud of designing a modular multi-agent architecture that brings multiple evidence types together in one workspace.

We also prioritized evidence provenance: original files remain preserved while extracted information, classifications, transcripts, web sources, and cross-references are stored separately.

Another accomplishment is incorporating audio transcription into the evidence workflow, allowing users to review timestamped transcripts alongside their original recordings.

What we learned

We learned that building an AI application involves much more than choosing a model and sending it a prompt. The surrounding architecture—including data processing, file handling, agent responsibilities, APIs, source tracking, and error handling—is equally important.

We also learned the importance of designing AI systems around uncertainty. AI-generated information should not automatically be treated as fact. Instead, results should remain traceable to their original evidence and sources, with humans responsible for interpreting the information.

What's next for Case Cabinet

Next, we want to improve the evidence cross-referencing system and make relationships between different pieces of evidence easier to visualize.

We also plan to expand file-format support, improve the accuracy of our specialized AI agents, strengthen source verification, and introduce video analysis capabilities.

Our long-term goal is to develop Case Cabinet into a reliable evidence organization and research workspace that helps investigators navigate complex information while keeping humans in control of how evidence is interpreted.