Aneeq KhatriAneeq Khatri / AI Engineer
AI systems. Built to hold up in the real world.
I build agent systems, retrieval pipelines, and evaluation tools that make model behavior easier to inspect.
- Agents
- Retrieval
- Evaluation
- Governance
My deepest expertise is finance and accounting. I worked in accounting before I started automating it.
AI for finance & accountingBased in Karachi, Pakistan. Building for teams everywhere.
agent.route(context)Give each agent a clear job.
Specialist agents share context, call tools, and hand work back to an orchestrator. Every handoff is traceable.
- 16AI projects built and evaluateddesigned, built and evaluated; none yet in production at a client
- 5 yrsindustry experiencethe last two building multi-agent and retrieval systems
- 5documented engineering case studiesarchitecture, evaluation and limitations available to inspect
- 5,000+engineers trainedacross three Pakistani national AI programmes
Recorded, narrated, unedited
Featured work.
1:42RevLedger
Tool-grounded answers over a deterministic subscription revenue engineWatch the walkthrough
1:34InvoiceAudit
Vision extraction checked by arithmetic, with bounded repair and a second check on the repair itselfWatch the walkthrough
1:48LedgerLens
An exception workbench where an uncited hypothesis is deleted in code, applied to the breaks AI matching cannot closeWatch the walkthrough
The demo is
the beginning.
A little more about me A model response is only one part of a product. I build the retrieval, evaluation, guardrails, and interfaces that make the whole system useful.
My work spans multi-agent coordination, document extraction, retrieval, and evaluation. An accounting background helps me identify where a plausible answer becomes a costly mistake. Here you can inspect implemented projects, trace their architecture, read measured results and limitations, and see how I separate deterministic checks from model judgment.
Before you ask
Questions people ask first.
Are you available for new work?
Yes. Open to fully remote roles worldwide, and to project work in the areas listed on the services page: agent engineering, evaluation and reliability, finance automation, and training. The form below or email is the quickest route.
Where are you based, and which time zone do you work in?
I am based in Karachi, Pakistan and work remotely, in any time zone: my hours follow the team I am working with.
What do you specialise in?
Agentic AI: designing, building, evaluating and operating multi-agent systems. That covers tool calling, retrieval, orchestration, context engineering and structured outputs, with evaluation pipelines, guardrails, human-in-the-loop controls and observability, so a system's behaviour is measured rather than assumed.
What makes your background different?
I worked in accounting before I moved into software, reconciling bank feeds against the ledger and closing the month. So I know where a plausible answer becomes a costly mistake, and I design agents around how finance teams actually work, with a person signing off where it matters.
Do you only work on finance?
No. Finance is my deepest domain, and its constraints are the strictest, but what I build transfers to any work that runs on documents, approvals and decisions: human gates at the risk boundaries, answers grounded in evidence, and evaluation that survives repetition.
What is your stack?
Python with LangGraph, LangChain, the OpenAI Agents SDK and MCP for the agents; TypeScript, React and Node.js for the product around them; and evaluation pipelines, guardrails, human-in-the-loop controls and observability on every system.
Do you also train teams?
Yes. I lead a faculty of around 100 instructors at the Governor Sindh Initiative and have trained thousands of engineers across Pakistan's national AI programmes. Training for teams and individuals is on the services page.
Open to fully remote roles worldwide
Have a hard problem?
Let’s build through it.
Tell me what you are building, where it gets difficult, and how I can help. You can also reach me directly by email.