All Models Great and Small
On why larger models are not always better
A unified legal data stack, AI and integration included.
Built for legal
Kelvin is for law firms, legal departments, alternative legal service providers, and legal technology providers that want a modern data stack built for their needs.
Law firms use Kelvin to turn their disconnected, unstructured data into data assets that create value for their clients and the firm's bottom line.
Alternative legal service providers use Kelvin to deliver data-driven legal process outsourcing, legal staffing, and legal consulting.
Legal Operations teams use Kelvin to create processes and capabilities that drive efficiency and value for their stakeholders.
Legal departments use Kelvin to standardize and enrich data throughout their systems, including contracts, matters, and invoices.
Legal technology providers use Kelvin to enhance legacy products and rapidly create new, data-driven legal technology products.
Legal data scientists use Kelvin to create data-driven legal dashboards and analytics in a fraction of the time.
Kelvin use cases
Financial data is the lifeblood of law firms and legal departments. Kelvin is used in pre-bill and post-bill by service providers and purchasers to support many common processes.
How long should a draft take? How should we allocate internal and external resources? Use Kelvin to create data-driven pricing and project management processes.
Automatically generate templates from your DMS. Populate contracts from sales emails. Triage inbound contracts to identify risks and key clauses. Kelvin allows you to rapidly prototype and build like never before.
Rapidly identify and organize documents from a variety of systems, extract key information, and present results to stakeholders. Kelvin changes the nature of post-execution review and diligence.
Ingest documents and data sources into a searchable index with built-in embedding vectors and extracted facts. Want documents from before 2021 involving Acme Corp and leasing concepts? Ask Kelvin.
Monitor sources like dockets or the Federal Register for specific clients, concepts, or events - connected directly to matter management system or DMS.
Legal domain knowledge
Language is the coin of the legal realm - but it sure isn't common currency outside. Kelvin ships language models built for legal and integrates with common LLMs like GPT, T5, Claude, PaLM, and BERT.
Summarize multiple documents, answer questions about contracts, build knowledge graphs, or execute agent-based research task - all within your firewall, integrated into your legal systems.
Kelvin natively speaks legal data formats you won't find elsewhere, like WordPerfect, LEDES, and EDRM-recommended formats.
Kelvin knows legal data models like SALI's LMSS, the ABA's UTBMS, and databases like Aderant and TeamConnect.
Kelvin knows private and public legal data sources like PACER, the Federal Register, EDGAR, UniCourt, and DocketAlarm.
For organizations interested in training their own models, the Kelvin Legal DataPack provides over 150B legal and financial tokens with safe commercial usage rights.
Step 1: Connect to data sources
Use Kelvin Share, DMS, Matter, Billing, or Research to connect to your data sources.
Step 2: Define a task
Use Kelvin's APIs or Python library to quickly define a task, like extracting metadata or finding anomalies.
Step 3: Execute at scale
Develop locally, execute remotely, deploy wherever. Kelvin runs on your MacBook Air or GPU clusters.
Step 4: Share results
Build dashboards or integrate data into existing systems using push or pull APIs or SQL.
On why larger models are not always better
273 Ventures CEO, Michael Bommarito, Quoted in LegalTech News Discussing Exaggerated Claims by Legal AI Vendors
A key component to enhancing AI's accuracy and reliability
Optimizing performance in the legal domain
FTC Tech Summit confirms importance of Kelvin Legal LLM
The size of a model does not uniformly define its success
A closer look at tokens and why they matter
A watershed moment for generative AI