I build AI applications with Python. I also develop their databases, APIs, and web interfaces.
I am an AI engineer in Brussels with seven years of experience. I work on language models, text processing, and computer vision.
I develop databases, backend services, and web interfaces. I also deploy applications and maintain them after release.
I work at ACA Group. I previously worked at Odoo and at Exellys, on projects for ING and Contract.Fit.
I discuss the current process with its users before I start development.
We agree on the tasks the software should handle. We clarify open questions during development.
I give users an early version so they can test it and suggest changes.
I design the database, build the application, and deploy it. I monitor it and fix problems after release.
I use OpenCode with GPT, Claude, and EU-hosted open-source models. The task and the client data rules determine which model I use.
I test search results and model responses. I also check data quality and handle failed requests.
I agree on a small first version with the client. I explain which features it includes and which can wait.
The bars show project dates on a shared scale. Scroll down for older projects. Select a project for details.
A language practice app with spoken conversations and corrections. I develop and run it through CodeBakery.
An application template for teams that need to process documents with AI.
A .NET service for municipal questions and incident reports. It uses submitted photos to classify incidents and set their priority.
Collects text from Belgian municipal websites for a question-answering service with source references.
A document system for personalized reports and chat. Answers include references to the source passages.
Uses language models to compare CVs with job descriptions and explain each score.
Answers questions using product manuals and links to the relevant sections.
A product I run through CodeBakery. It compares candidates with job descriptions and explains the scores.
Plans walking routes that favor safe crossings. A language model writes spoken directions for each route.
Extracts text from scanned case files, separates the documents, and makes them searchable by meaning.
A dashboard for country-risk analysts. It uses language models to analyze documents and news articles.
Detects objects on Jetson Orin devices in garbage trucks. AWS Greengrass manages the devices.
A service template with authentication, caching, messaging, and database connections.
Python data processing on a shared Kubernetes platform for Flemish government departments.
A Django application for field appointments. It uses existing data from Navision and Windows Active Directory.
A horoscope generator using a fine-tuned GPT-Neo model. A FastAPI service provides predictions.
Collects product information from e-commerce search results. A Flask API provides access to the results.
My software consultancy. I use it for client projects and for my products, Speakroom and Swiftalent.
Features and performance improvements for Odoo websites and online shops, using Python, JavaScript, and OWL.
Migration of machine learning models and data processing to Google Cloud, with automated deployment and scheduled runs.
Classifies documents and extracts fields from invoices and receipts. It includes Dutch text processing.
An EDSR model that increases image resolution.
Research on document classification using both images and text. I evaluated the models on RVL-CDIP and Tobacco3482.
You can contact me about AI applications, backend services, or internal tools.
EMAIL[email protected] LINKEDINin/kamen-zhekov ↗