Kıvanç Yüksel
From: Poland
Location: Warsaw
On Useme since 8 October 2026
About me
Kıvanç is a machine learning engineer with several years of experience building AI products that real people use. For four years at Allegro, one of Europe's largest marketplaces, he worked on visual search that lets shoppers find products from a photo, and his team's improvements increased sales. Today, he builds generative AI applications: chatbots, AI assistants, and image and video generation. He can take an idea from the first prototype to a finished, reliable product.
Categories
CV / Résumé
Jan 2026 - Now
Founder and AI Engineer
SmileToAI
Founded and built SmileToAI alone: a creative AI platform that turns a single prompt into images, video, narration, and answers from an AI assistant, on web and mobile.
Dec 2021 - Dec 2025
Machine Learning Research Engineer
Allegro
Worked on visual search end-to-end, from research and dataset construction to training, evaluation, and production. Led monthly knowledge-sharing seminars for 4-8 engineers and cross-team sessions every 3-4 months.
Oct 2021 - Dec 2024
Machine Learning Engineer
10Clouds
Owned the full lifecycle of each engagement: data collection, labeling and versioning, model selection and training, evaluation, serving, and handover to DevOps.
Nov 2018 - Oct 2021
Machine Learning Engineer
Promity
Built the ML core of a real-time video analytics system for retail stores, combining object detection and tracking, face detection and recognition, facial landmarks, emotion, age, gender estimation, and motion detection in one live pipeline.
Portfolio
A creative AI platform I designed and built alone that turns a single prompt into images, video, narration, and answers from an AI assistant. It runs on the web, iOS, and Android and is powered by models from multiple providers
Real-time Video Analytics for Retail Stores
A live video analytics system that helps retail businesses understand their customers from in-store cameras. I built its machine learning core.
My master's research at Warsaw University of Technology on detecting facial landmarks, the key points of a face such as the eyes, nose, and mouth, using convolutional and recurrent neural networks.
Kıvanç built SmileToAI's hands-free voice mode, which is live with paying users on the web and mobile. Users talk to the assistant in real time and can interrupt it mid-answer. It runs on OpenAI's Realtime API over WebRTC.
Machine Learning Engineer
Visual search for Allegro, one of Europe's largest online marketplaces, which lets shoppers find products by searching with a photo. The system searches a catalog of over 150 million items, handling 1,600 queries per second in under a second each.
An open-source tool that speeds up video editing by detecting silent parts of a video and cutting them out automatically. It's useful for lectures, tutorials, and recordings where pauses make up a large part of the footage.
Inside SmileToAI, teams upload PDFs, Word files, slides, and spreadsheets, and ask questions about them. Kıvanç built the full pipeline: documents are split by section, indexed for both meaning-based search and a graph of the people, products, ...