Arash ASKARI

Enterprise · Cloud · AI architect · Zürich

Agentic AI with proof, control, and an owner.

I design and test AI systems for regulated companies. The work covers finance workflows, cloud platforms, data, and governance.

Discuss a system problem ↗ See the work →
Workstation with an Azure enterprise AI architecture, deployment flow, security key, and compact computer

Practice areas

  1. 01 FRAME Agentic systems
  2. 02 TRANSFORM Enterprise transformation
  3. 03 GROUND Cloud + data

Selected work · Agentic AI

Agentic finance systems for a regulated company

I designed one reference architecture and built four finance POCs: contract capture, invoice chat, document links, and billing checks.

  • Four POCs use real finance documents and workflows
  • One reference architecture supports all four agents
  • A person can trace each invoice result to its source document

System Azure · Reflex · LangChain · LLM agents · MCP

Read the case →

Operating model

Frame. Prove. Govern.

  1. 01

    FRAME

    Name the decision before you select the tool.

    Define the outcome, evidence, limits, and owner before you discuss a model or cloud service.

  2. 02

    PROVE

    Build the smallest proof that can fail.

    A running workflow shows real behavior, failure points, and operating cost. A slide deck does not.

  3. 03

    GOVERN

    Put control inside the design.

    Add human approval, evaluation, logs, and named ownership to the system. Do not leave them for a compliance review.

Project index

Systems, decisions, and results

View all work →

How I work

I stay close to the system.

I work with leaders and delivery teams in regulated companies. I test the hard assumption, build the proof, and name the owner.

How I work →

A hard system problem?

Start with the decision. Then prove it.

Send me the decision, constraint, or failing system. I’ll tell you where I would start.

Send the problem ↗ View work →