End-to-end development of multimodal embedding models
Leading the Face Recognition team
Full SOTA face recognition pipeline (from classical CV to deep learning):
Face detection, alignment, generation, quality, embedding extraction, data scrapping and processing, training on large-scale datasets.
Achievements and challenges
1st place NIST FRVT April, 2021 and September, 2023 - 2024
Gold medal in Happywhale - Whale and Dolphin Identification challenge
LLM stack
Development of AI agents utilizing Chain of Thought (CoT), structured outputs, functional calling, Retrieval-Augmented Generation (RAG), etc.
Work experience
1) Classical computer vision: edge detection, viola-jones algorithm, keypoints detestion (SURF, etc), segmentation (GrabCut), tracking, etc.
1) Image classification. Painting classification (style, genre), Type of room, etc.
2) Liveness (anti-spoofing), commercial product. Also took 2nd place in a face anti-spoofing challenge
3) Several projects with Microsoft HoloLens, also gesture recognition with MS Kinect
Main publications
1) Automatic coloring of grayscale images based on intelligent scene analysis (https://link.springer.com/article/10.1134/S1054661815010022)
2) Context-Sensitive Image Analysis for Coloring Nature Images
(https://link.springer.com/chapter/10.1007/978-3-319-33816-3_14)
Master's degree
PhD studies
Upper-Intermediate
https://habr.com/users/Kwent/
https://www.kaggle.com/kwentar
https://www.linkedin.com/in/aleksey-alekseev/