Freelance AI engineer for documents, automation and AI agents
If your team spends its days retyping PDFs, reading scans or answering the same questions, I can probably help. I build AI that reads documents and images and does the boring steps for you, and I build it for real use, not for a demo.
Intelligent document processing
PDFs, invoices and orders turned into clean data, no retyping.
- Sound familiar?
- Someone on your team copies numbers from PDFs, invoices or order emails into another system. Every day.
- What I build
- An AI pipeline that reads each document, pulls out the fields you care about (totals, line items, dates, customer details), checks them and sends them where they need to go: your ERP, a spreadsheet, an API.
- What you get
- The typing is gone, and a person only looks at the few documents the AI isn't sure about.
Typical uses
- Invoice data extraction
- Purchase orders from email
- Receipts and bank statements
- Contracts and delivery notes
- PDF to structured data
- Invoice extraction
- OCR + LLM
- JSON output
Real project: Order extraction from emails and PDFs →
OCR for scans and handwriting
Scanned forms and handwritten notes, read properly.
- Sound familiar?
- You have boxes (or folders) of scanned forms and handwritten pages that nobody can search.
- What I build
- OCR that goes past plain text: it understands the layout, reads handwriting, ticks and stamps, and fills a proper form or database record for every page.
- What you get
- Paper archives you can search, filter and use, instead of paper archives you avoid.
Typical uses
- Handwritten forms
- Scanned archives
- Field reports
- Stamped and signed documents
- Handwriting recognition
- Layout-aware OCR
- Document digitization
Real project: Handwritten form digitizer →
AI agents and workflow automation
AI that doesn't just answer, it does the next step too.
- Sound familiar?
- A process has many small steps: read the request, check it against the rules, look something up, route it, reply.
- What I build
- AI agents that run those steps for you, using your own tools and data (email, CRM, databases, APIs), with a human approving anything important.
- What you get
- Requests handled in minutes instead of days, with a clear log of what the AI did and why.
Typical uses
- Claims and application processing
- Email triage and replies
- Checks against your rules
- Hand-off to the right person
- Agentic AI
- LLM tool use
- Human in the loop
- Process automation
Real project: Insurance claim processing →
AI assistant for your company documents (RAG)
Ask your manuals, policies and reports a question, get an answer with sources.
- Sound familiar?
- The answer exists somewhere in your documents, but finding it takes half an hour and three colleagues.
- What I build
- A private chatbot built on retrieval-augmented generation (RAG): it searches your own documents, answers in plain language and shows exactly which page the answer came from.
- What you get
- New people get up to speed faster, and experts stop answering the same questions.
Typical uses
- Internal knowledge base
- Technical manuals
- Policies and procedures
- Customer support answers
- RAG
- Semantic search
- Answers with citations
- Private LLM
Real project: Company knowledge assistant with sources →
Computer vision and image analysis
Photos, labels and drawings checked by AI.
- Sound familiar?
- People spend hours looking at photos, product labels or technical drawings to check the same things.
- What I build
- Vision models that look at each image or drawing and report what matters: missing label information, progress on a building site, rooms and sizes on a floor plan.
- What you get
- Faster checks that are the same every time, with the doubtful cases flagged for a person.
Typical uses
- Label compliance checks
- Floor plans and technical drawings
- Construction progress photos
- Product photos
- Image recognition
- Vision LLM
- Quality control
Real project: Pet food label compliance checks →
LLM integration into your software
GPT, Claude or Gemini features inside the app you already have.
- Sound familiar?
- You'd like AI inside your product or internal tool, but don't want a science project.
- What I build
- A well-tested AI feature added to your existing app: the right model for the job, prompts that hold up with real data, costs kept in check, and a fallback when the AI isn't sure.
- What you get
- An AI feature your users actually use, not a demo that breaks on the first real customer.
Typical uses
- Smart search
- Summaries and drafts
- Classification and tagging
- AI-powered web apps
- OpenAI API
- Claude API
- Python
- Next.js
Real project: Receipt parser →
How we'd work together
Step 1
A quick call
15 minutes, free. You show me the documents or the process, I tell you honestly if AI is a good fit.
Step 2
A small test on your data
Before anything big, I run a sample of your real documents through a first version, so we both see how well it works.
Step 3
The build
I build it, connect it to your systems and test it with the people who'll use it. You see progress every week.
Step 4
Live, and looked after
It goes live, I keep an eye on it, and it keeps improving as it sees more of your documents.
Why people work with me
- 50+ AI projects built, most of them on messy real-world documents
- You talk to the engineer who builds it, no account managers
- 1st place at the SheepAI hackathon in Zagreb, 2025
- Honest answers, including "AI isn't the right tool for this"
Questions people usually ask
+Do you work as a freelancer or through a company?
Both, depending on the size. You always work with me directly. Bigger and ongoing projects go through MindX Global, my AI company.
+What does a project cost?
It depends on how many document types and systems are involved. After the first call you get a written scope and a price before any work starts.
+Is my data safe?
Your documents are only used for your project. When needed, the system can run on your own servers or in the cloud region you choose.
+Do I need clean data or a data team?
No. Messy scans, mixed PDFs and handwriting are the normal case. You just need someone who knows the process and can answer questions.
+Where are you based?
Varaždin, Croatia. I work remotely, with companies in Europe and further away, in English or Croatian.