In development · finetuning
docket.ai
AI intake automation for law firms. Captures intake calls, qualifies matters, drafts engagement letters, and routes to attorney review.
Production AI systems. Built to ship.
Bespoke systems for clients who need the model behind a real workflow, not a slide. Multi-model infrastructure, agentic systems, and the evaluation discipline that keeps them honest in production.
Shipped & in-flight
Operational — productizing
the-array
The operational AI infrastructure behind the practice, productized: a LiteLLM gateway routing across cloud and local models, with full observability and a maintained model catalog.
Capabilities
Praxen runs one AI stack across the hardware on hand: phone, desktop, an always-on workstation, and the cloud behind them. One gateway, the-array, routes every model call across the fleet. Here is what that gets you.
Every model, one door
Claude, GPT, Grok, Gemini, and models running on local hardware, all behind one endpoint and billed per use under a ceiling you set.
It never goes dark
When a provider has an outage, the work reroutes to the next model automatically, down to a model on local hardware. A single vendor going down costs you nothing.
Private work stays private
Sensitive material can run on hardware in the building and never reach a third-party cloud. That is the routing table, not a policy line.
Every dollar on a dashboard
Each call is priced and logged, and a hard monthly ceiling stops spend before it happens. The bill never surprises you.
No lock-in
Switch vendors with a config change instead of a migration. The model behind your workflow is one you can pin to a version and re-run.
Built, not bought
This is the production system Praxen runs in-house, offered to clients. Twenty-one years of hardware engineering behind it, and the stack itself is the proof.
Talking to us
For client engagements: chris@cld-dev.io. Include the problem, the constraint, and what shipping looks like for you.