Corpus Law AI

Built by lawyers and engineers who understand both worlds.

Preceda is the flagship product of Corpus Law AI — a boutique AI technology company building intelligent tools for specialized knowledge domains. We started with law because it is where the gap between what technology can do and what practitioners actually have is the widest.

The team

Our founding team brings together deep legal practice and engineering at scale. Our legal co-founder has spent 35 years practicing before the Delhi High Court and the Supreme Court of India — he knows what lawyers actually need, not what technologists assume they need. He has litigated complex matters across constitutional, commercial, and criminal law, and brings the kind of domain expertise that cannot be replicated by reading textbooks or scraping databases.

On the technology side, our engineers are veterans of building ML training infrastructure, large-scale retrieval systems, and production AI platforms. They have designed systems that serve millions of users and trained models from scratch — not fine-tuned wrappers around someone else's API, but ground-up architectures built for the specific demands of legal text.

The problem

Legal research in India has long relied on databases that prioritize volume over accuracy. Citations are often unverified. Case names are garbled from OCR artifacts. Metadata is incomplete or wrong. Lawyers spend hours cross-checking references that should be trustworthy out of the box.

Meanwhile, the tools available in Western jurisdictions are priced for large firms and built for common law systems that do not map cleanly to Indian legal practice. There is no single platform where a lawyer in Delhi can search Indian case law alongside Canadian or UK precedent with the same confidence in citation accuracy.

What we built

Preceda is a jurisdiction-first legal intelligence platform covering India, Canada, the US, and the UK. The corpus is comprehensive, with a verified citation graph where every reference between cases is checked against the actual corpus, classified by treatment, and scored for authority.

The AI matter agent drafts court-ready legal documents with 22 specialized tools — searching case law, verifying authorities, checking limitation periods, computing court fees, analyzing judge patterns, and producing bail applications, writs, appeals, and legal notices grounded in real case law.

Intelligence profiles for over 30,000 lawyers and 22,000 judges provide win rates, bail grant rates, citation authority scores, and disposition speed — derived entirely from case data, not self-reported.

The technology

We built the first retrieval-optimized embedding model for Indian law — trained on 637,000+ citation-graph-derived pairs across 13 Indian citation formats. The model runs locally via ONNX at zero marginal cost, with no API dependency.

Citation resolution uses deterministic patterns across 40+ jurisdiction-specific formats, not language model guessing. Treatment classification covers 366,000+ edges — followed, distinguished, overruled, applied. PageRank authority scoring ensures landmark decisions surface above semantically similar but obscure orders.

The entire system is built for legal accuracy at scale: verified citations, deduplicated corpora, repealed law detection, and OCR artifact correction across 3.5 million cases.

Corpus Law AI

Preceda is our first product. The citation graph and the intelligence layer we have built are the foundation for a family of tools that bring AI to specialized professional domains — starting with law, where accuracy is not optional and hallucination is malpractice.