Launched this week

Casero

Launched this week

A better legal memory for your law firm

164 followers

A better legal memory for your law firm

164 followers

Visit website

•

AI Infrastructure Tools

•

LLM Memory

Law firms are sitting on decades of winning work but every lawyer still starts from scratch. Casero turns a firm's existing cases, emails, and documents into a living knowledge graph. Every AI answer links back to the exact source passage. Closed cases become reusable precedent, matched by facts and legislation, not keywords. And the system flags blocked deadlines before they become problems. Most legal AI is a smarter search box. Casero is institutional memory.

Free Options

Launch tags:Legal•Tech•YC Application

Launch Team / Built With

Framer — Launch websites with enterprise needs at startup speeds.

Launch websites with enterprise needs at startup speeds.

Promoted

Casero

Maker

📌

We built Casero after seeing how legal work is actually managed day to day inside firms. Most matters live across scattered documents, emails, notes, and folders. Lawyers constantly have to reconstruct context, chase missing information, and search through files just to understand what is going on in a case. Existing systems mostly store files and timelines, but they don’t actually understand the matter itself. So we asked a simple question: What if the system actually understood the matter? Casero automatically pulls information from documents, emails and notes, extracts key facts, and keeps a live structured summary of the case. It connects everything happening in a matter so lawyers can instantly understand the state of the case, find information, and spot missing steps or risks. The goal is simple: stop lawyers from spending hours rebuilding context and let the system maintain it automatically. We would genuinely love your thoughts and feedback. If you work in legal or know someone who does, we would really appreciate hearing how this could be useful or what you would want improved. Thanks for checking it out 🙏

Report

1d ago

@moeed_chughtai Hey Moeed — the context reconstruction problem is real. The hardest technical challenge here is probably relevance ranking when a lawyer queries across hundreds of documents.

Curious — how are you handling semantic search? Pure vector similarity or are you using a reranker on top to improve precision? BGE-M3 cross-encoder reranking could significantly improve your retrieval accuracy for legal documents specifically.

Report

2h ago

Casero

Maker

@kunal_raha Yes you're absolutely correct about the technical challenges. We are currently using a hybrid approach of a few techniques. As this is just the MVP, we are continuously looking at different ways to approach this and benchmarking against each other. Appreciate your input for the reranking!

Report

2h ago

@moeed_chughtai That's great to hear! I've actually built a BGE-M3 semantic search microservice optimized for CPU — ~1s latency, no GPU required, Dockerized with ChromaDB. Designed for exactly this kind of document retrieval use case.

Here's the repo if it helps with benchmarking: github.com/kunalraha-ai/Optimized-Search-Engine-Reranker

Curious to hear how it compares against your current approach

Report

1h ago