Projects · completed jobs

Systems I shipped

Every project below runs in a real company today. Each one started as a specific manual task that someone on the team did by hand, week after week.

B2B equipment supplier

AI sales automation assistant

Creating customer orders in the ERP was the team's most time-consuming daily routine. An AI agent now drafts the invoice from the incoming request, and managers verify instead of retyping.

hand-typed→auto-drafted

customer invoices

customer request ai agent draft invoice manager review

Jewelry retailer

AI catalogue image standardizer

Supplier photos arrived in chaotic sizes, crops, and orientations, and 70% of shoppers browse on mobile. An AI service now normalizes every product photo to the catalogue standard, inside the existing publishing pipeline.

chaos→uniform

catalogue of 8,000 photos

product photo ai vision pre-filter ai image editing quality check online store

Grant consultancy

AI grant application expert

Drafting one application took an industry expert several weeks. The system now reads the company's documents, researches the market, and writes the draft in Google Docs, and the expert reviews and comments instead of writing.

weeks→hours

per application draft

company documents ai agent draft in google docs expert review

Event-management platform

AI assistant for event management

Managers spent up to 3 hours assembling each event budget by hand. An AI assistant embedded in the platform now drafts it in minutes, and the manager reviews instead of copy-pasting.

3 hrs→minutes

per project budget

project brief ai assistant budget draft manager review

Ad platform · ~200 clients

AI support and onboarding agents for an ad platform

Every signup needed hand-holding from the team. AI agents now answer support and onboarding questions from the internal knowledge base and the client's own CRM data.

manual→self-serve

onboarding, ~200 clients

client question ai agent answer in chat

Services company

Self-updating CRM in Google Sheets

Customer emails landed in Gmail and the CRM lived in Google Sheets, updated by hand. An LLM now reads the correspondence and fills in status, deadline, and the latest agreement on its own.

inbox→crm rows

runs every 5 min

gmail inbox client match ai analysis sheets crm

Content platform

Screenshot quality assessment

Checking large volumes of screenshots for text legibility by hand didn't scale. A microservice now scores each image with OCR and image analysis and returns the rating through a REST API.

manual review→api score

good / fair / poor

screenshot task queue ai analysis quality score

Got a process like these?

Book a 30-min discovery call. I'll tell you honestly if AI is the right fix.

Book a 30-min discovery call