Applied AI systems

Working AI for document-heavy operations.

One repetitive workflow at a time, built into the systems and permissions you already have, with a person still signing off.

Discuss a workflow

What we build

Six things we build.

Start with the one that costs you the most hours.

Document processing

Recurring PDFs, forms and spreadsheets become structured records, each field linked to its source.

Internal knowledge

Answers from your own policies and records, with the source cited and the permission rules intact.

Voice agents

Phone workflows with a clear script, such as intake and scheduling, with a clean handoff when a call gets complicated.

Workflow automation

Gathers the context, prepares the next action, and routes it through the tools your team already uses.

Evaluations and reliability

Define what a good result is, test it on realistic cases, and keep testing as data, prompts and models change.

Systems integration

Wire the model into the records, screens and controls it needs to be useful in a real operation.

See what we build

Industry specialization

Applied AI for fund administration.

Pranav ran fund-administration operations before building AI for them: NAV close, reconciliation breaks, investor requests, document intake.

Explore fund administration

Built for production

The model reads. Plain software decides.

Permissions, approvals and record updates stay in deterministic code. Before anyone relies on a workflow, we test it on the cases it will actually meet.

How we evaluate AI systems

Direct delivery

Work directly with the person building the system.

Velug is one person, Pranav Reddy. You talk to the person who writes the code, from the first workflow map to the last test.

About Velug

Start with one workflow

Which process eats the most hours?

Email a rough description of it. I reply myself.

Email Pranav