AI engineering
AI engineering: AI workflows, agents, and applications
I build software that uses large language models: AI features in an application you already run, agents and workflows that take on repetitive work, and standalone AI products. I connect them to your data and systems, test them on examples from your work, and keep them running after launch. The model is one part of the job. The rest is software engineering, and that's what I've done for my whole career.
Work directly with Tim KamaninIndependent developer since 2007
Who this service is for
You might be a founder building an AI product, a team adding AI features to an existing application, or an agency that needs a developer. You might have a process you'd like to automate, or a prototype that works in a demo but isn't ready for real use. Bring an idea, or we can start from scratch: you show me how your team works today, and we find what's worth automating.
How the service works
For example
You run a customer portal, and your support team answers the same kinds of questions every day. The answers are in help articles, product documents, and past tickets, but finding them takes time. I build a workflow that takes each customer inquiry, searches that material, and drafts a reply with links to its sources for your team to check and send.
This is an example. We'll work out the details for your project together.
We choose a problem worth solving with AI
You show me the work or the product idea. We decide where an AI model helps and where ordinary code is simpler and more reliable. Then we pick a first use case and collect real examples, each with what a good result looks like.
What you get: A first use case and a set of examples to measure results against.
I build a working prototype on your examples
I build the core feature with your data and test it against the examples. You try it, we review its output together, and I adjust the instructions, the information it draws on, and the code around the model.
What you get: A prototype you can try on your own material, and an early answer on whether the approach works.
I build it into your software
I connect the data and systems it needs, such as documents, customer records, email, or your existing application. I add permissions, review steps, error handling, and the interface around it.
What you get: An AI workflow or agent working inside your systems, with your team in control of its actions.
I test it before your team relies on it
I turn real cases into a test set and measure accuracy, response time, and running costs. I also try unusual inputs and look for the ways it can fail. Your team reviews the results, I fix what falls short, and together we decide when it's ready for everyday use.
What you get: Measured results on your own tasks, so you know what it does well and where it needs a person to check.
You use it every day, and I keep it on track
I deploy it and walk your team through it. I watch quality and costs, rerun the tests when models or your data change, and build the next improvements with you.
What you get: An AI system in daily use, with continued support from the developer who built it.
What you get
- The AI workflow, agent, or application we've planned together.
- Connections to your data and business systems.
- Permissions and review steps that keep your team in control of its actions.
- A test set built from your examples, so you can measure changes over time.
- Deployment, monitoring, documentation, and a walkthrough for your team.
- A plan for maintenance and further improvements.
How we’ll work together
You explain the work and provide examples; I build the software. We agree on what it should do, the budget, and how we'll keep in touch. Before we build, I explain the expected running costs and how I'll keep the system working as models change.
Relevant experience
I've worked independently since 2007, building and maintaining web applications. AI systems depend on the same work: data, permissions, integrations, testing, deployment, and maintenance.
More about my workQuestions about AI engineering
What does AI engineering involve?
It's the work of turning AI models into dependable software. The model is one part. The rest is connecting it to your data, deciding what it's allowed to do, testing its output, handling failures, keeping costs predictable, and maintaining it after launch.
What can a custom agent do?
An agent can find information, research a question, or work through a multi-step task using the systems you connect to it. We define what it can access, which actions it can take, and where your team reviews its work.
Which AI models do you use?
We choose based on your task, data, budget, and privacy requirements. I test the options on your examples and build the system so you can switch models later without starting over.
How do we know the results are good enough?
We build a set of test cases from real examples and agree on what a good result looks like. I measure the system against them before launch and after each change, and we add review steps where mistakes would be costly.
Can my team review the AI's work before it takes action?
Yes. We decide where review is useful, for example before sending a reply or changing a customer record. I build those checkpoints into the process so your team can approve or correct the proposed action.
Can you improve an existing AI prototype?
Yes. Many prototypes work in a demo but struggle with real data. I review what you've built, test it on examples from your work, and fix what it needs to hold up in daily use with more users.
AI engineering
Tell me what you need
Describe the product, feature, or task you have in mind, and the systems or data involved. If you already have a prototype, tell me what it does and where it falls short. A few sentences are enough to start.
I'll read your message and reply with questions or a suggested next step.