AI integration for companies and existing products
AI is worth adding when it solves a concrete problem: someone retypes documents by hand, customers cannot find the right offer, or the team wastes time looking for information. I add features like that to an existing product or process, without rewriting everything.
Who it is for
- Companies with a live product that want an AI feature for their customers.
- Teams with many documents and internal knowledge they want to use faster.
- Products where personalization or recommendations can change what users see.
What I can do
- Choosing where AI has a measurable effect and how we will measure it.
- An assistant with access to company context (RAG) and sources in its answers.
- OCR and data extraction from documents.
- Recommendations, personalization, and classification.
- Quality monitoring: traces, evals, and model costs.
How I work
01
Workshop
We review the product and processes and pick one feature with the best effort-to-impact ratio.
02
Scope and metric
We agree on what the feature should do and how we will know it works.
03
Two-week sprints
First the feature runs on real data in a limited scope. Then I wire it into the product: API, UI, permissions, logs.
04
Clickable version
You can try the feature inside your product as early as possible.
05
Measurement and iterations
We check that it does what it should and decide what to improve in the next sprint.
- I do not give fixed timelines, because they depend on the size of the app. After every sprint you see what is done and what is left.
- I bill hourly and invoice as a B2B contractor.
- I design the UX and UI myself, alongside the code.
- I reply to messages within 24 hours.
Projects that back it up
Foodtech platform
I led AI initiatives end to end for one of the largest foodtech companies in Poland: an AI assistant, recommendations, and personalization on mobile and web.
See project: Foodtech platformCarSense
AI extracts data from car documents and turns it into specific tasks and reminders.
See project: CarSensecmdFlow
An on-device model on the Mac with automatic fallback to the cloud where the local model falls short.
See project: cmdFlowTechnologies
FAQ
Where should a company start with AI?
With one process or feature that has a clear goal, not a strategy for everything. One working feature teaches a company more than a long presentation.
Will our data go to OpenAI?
That depends on the model choice. We can use cloud models with the right settings, models hosted in Europe, or local models. We decide together before any code is written.
What is RAG?
RAG is how a model answers from your documents: it first retrieves matching passages, then answers using them and can point to the source.
How do we know the AI works?
Before rollout we agree on a metric, such as handling time, number of manual corrections, or conversion. Then a set of test cases checks every change.
How do we work together and how do you bill?
We start with a workshop where we pick the feature and the metric. Then I work in two-week sprints, and after each one you see a working piece in your product. I bill hourly on a B2B invoice.
Read next
Tell me what you want to build
A few sentences are enough: what problem you solve, for whom, and what you already have. I reply within 24 hours with questions or a workshop proposal.