AI features that answer from your data
Most AI features are a chat box wired to a model, and they guess. I build the part that makes them useful: retrieval over your own data, tools the model can actually call, and the guardrails that keep an agent inside the job it was given. Shipped in production, not a demo.
What I build
RAG, tool calling and AI agents, inside the product your users already have.
- RAG pipelines over your documents and database, so every answer is grounded in your records and can cite where it came from.
- Tool and function calling that lets an assistant read live data and take real actions — book the slot, update the record, start the workflow.
- Agentic workflows that plan across several steps and know when to stop, hand off or ask a person.
- Streaming chat UIs with structured output, so the interface renders real components instead of a wall of text.
- Evals and tracing, because an AI feature nobody measures quietly gets worse.
Built with Vercel AI SDK, LangChain.js, LangGraph.js, TanStack AI, Gemini API, OpenAI and Anthropic APIs, Vector search, MCP.
Working together
Full-time, contract or a single project, from Lahore to wherever your team is.
- Remote full-time
- For companies hiring a senior engineer or team lead.
- I join your team, take ownership of features end to end and help the people around me ship faster.
- Part-time or contract
- For teams that need extra hands for a few months.
- A set number of hours each week on your product, working inside your tools, standups and sprints.
- For agencies
- For dev shops with more client work than their team can take.
- White-label is fine: I work under your name, sign your NDA before kickoff and the code is your client's. Blocks of 20 to 40 hours a week, with 4 to 5 hours of overlap with the UK and EU, or US East mornings.
- Freelance projects
- For founders and businesses with a product to build.
- Clear scope, milestones and regular demos, from the first screen to the final deploy.
- Quick fixes
- For one-off bugs and small tasks.
- Tell me what's wrong and I'll give you a clear estimate before I touch anything.
What it costs
It depends on the scope, so there's no rate card. After a free 15-minute call you get the scope, timeline and cost in writing, before any work starts. Small fixes get a clear estimate up front too.
How a project goes
A free 15-minute call
We talk about what you need. No pressure and no sales pitch.
A clear plan
You get the scope, timeline and cost in writing before any work starts.
Build and demo
I work in short sprints and show you real progress along the way.
Launch and support
We ship it, and I stick around to fix and improve things after.
Questions
Can you build on our own data?
Yes, that's the point of the work. Retrieval runs over your documents and database, so answers come from your records and can cite where they came from. On OWNRS UAE the assistant answers from each owner's own property data.
How do you stop the AI from making things up?
By not letting it answer from memory alone. Retrieval hands the model the relevant records, tool calling lets it read live data instead of guessing, and the workflow keeps an agent inside the job it was given. Evals and tracing show when quality drifts.
Can it run on our own servers?
Yes. AI See runs on-premises on a client's industrial machines, reachable only over Tailscale VPN.
Do you only do the AI part?
No. I also build the web app, the mobile app and the backend the AI feature lives in, and deploy it, so you don't need three people to ship one thing.
Which models do you work with?
I've shipped on the Gemini API, and I work with the OpenAI and Anthropic APIs too, through the Vercel AI SDK, LangChain.js or LangGraph.js depending on what the project needs.
Where are you, and do time zones get in the way?
I'm in Lahore, Pakistan (UTC+5). I've worked with teams in Saudi Arabia, the UAE and the US, so remote work across time zones isn't new to me.
Tell me what you're building
A half-formed idea is enough. I'll tell you whether I'm the right person, roughly what it would take, and what I'd do first.