HT

Computer vision & robotics

Hakob
Tamazyan.

I train vision models, develop aerial navigation methods, and build machine learning systems that run in real time.

Download CV Email me 

01 / Selected work

From visual representations
to action.

01Foundation models

χViT & GeoCrossBench

I trained χViT, a self-supervised vision foundation model for remote sensing, entirely from scratch across multiple nodes. As first author of GeoCrossBench, I study how vision models generalize across satellite bands.

Self-supervised pretraining · Distributed training · Cross-band evaluation

Paper Code Model weights 
02Aerial robotics

Navigation with maps and memory

I develop and modify multimodal foundation models for drone navigation as vision-language-action models. My work combines map grounding, trajectory history, and simulation-generated data for instruction-conditioned action prediction.

Vision-language-action models · History-aware navigation · Synthetic training data

Research at YerevaNN 
03Real-time vision

Video segmentation & matting

At Krisp, I co-led the development of real-time video segmentation for background replacement and built video matting models from scratch. I designed the project’s ML architecture across model development, training, and inference.

Foreground estimation · Post-training quantization · Quantization-aware training

02 / Publications

Research in the open.

Google Scholar 
2026

ShipCross-ID: Bridging SAR and Multispectral Imagery for Ship Re-Identification

ECCV · TerraBytes II Workshop

2026

Beyond Data Size: Exploring the Impact of Dataset Diversity and Density in Self-Distillation Learning

ICLR · DATA-FM Workshop

Six more publications · 2023–2025
2025

Less is More? Data Specialization for Self-Supervised Remote Sensing Models

ICML · DataWorld and TerraBytes Workshops

2025

Teaching Visual Language Models to Navigate Using Maps

ICLR · Robot Learning Workshop

2024

Benchmarking Robustness of Foundation Models for Remote Sensing

ECCV · OODCV Workshop

2024

A Hierarchy of Determinative Sequent Systems with Different Substitution Rules

Pattern Recognition and Image Analysis

2023

The Relationship Between the Proof Complexities of Linear Proofs in Quantified Sequent Calculus and Substitution Frege Systems

Mathematical Problems of Computer Science

03 / Experience

Hands-on work.
Technical leadership.

Apr 2023 – Present

YerevaNN Lab

Computer Vision and Robotics Team Lead

  • Lead vision and robotics teams, combining model development with research supervision.
  • Train self-supervised foundation models and develop multimodal models for aerial navigation.
  • Research cross-band generalization, map-grounded navigation, foundation-model evaluation, and cross-modal retrieval.
Nov 2022 – Mar 2023

Mobeus

Staff Machine Learning Engineer

Developed real-time recognition methods for static and dynamic hand gestures.

May 2019 – Nov 2022

Krisp

Staff Machine Learning EngineerJan 2022 – Nov 2022
Senior Machine Learning Engineer IIJan 2020 – Dec 2021
Machine Learning Scientist/EngineerMay 2019 – Dec 2019
  • Co-led real-time video segmentation for background replacement and built video matting models from scratch.
  • Designed the ML architecture across development, training, and inference.
  • Optimized models using post-training quantization and quantization-aware training.
  • Contributed to audio dereverberation and noise-cancellation research for speech enhancement.
Sep 2017 – Apr 2019

BetConstruct

Machine Learning Engineer

Built models for time-series forecasting, image classification and clustering, football-match object detection, and user-behaviour prediction.

Nov 2016 – Apr 2017

Mentor Graphics

C++ Software Engineer Intern

Completed a software-engineering internship focused on C++ development.

04 / Background

Mathematical foundations.

PhD in Applied Mathematics

Yerevan State University · 2021–2024
Specialization: Mathematical Logic

MSc in Discrete Mathematics & Theoretical Computer Science

Yerevan State University · 2019–2021
GPA 3.95/4.00

BSc in Informatics & Applied Mathematics

Yerevan State University · 2015–2019
GPA 3.90/4.00

Technical toolkit.

Programming
Python, C++
Frameworks
PyTorch, TensorFlow, Keras, OpenCV, scikit-learn
Training & research
Multi-node distributed training, self-supervised pretraining, multimodal learning, vision-language-action models, aerial navigation, model evaluation
Deployment
Real-time computer vision, segmentation and matting, quantization, QAT

Academic activities

  • Program Lead — Vision-Language Foundation Models for Aerial Robotics
  • ICLR 2026 Peer Reviewer
  • Teaching Assistant, Mathematical Logic — Yerevan State University, 2021–2022
  • ICML 2025 Spotlight — Oral presentation
  • DataFest Yerevan Speaker — 2020, 2024

Honours & awards

  • International Mathematical Olympiad — Silver Medal, 2015; Honourable Mention, 2014
  • Zhautykov Mathematics Olympiad — Silver Medal, 2015
  • ICPC NEERC — Third Diploma, 2016, 2019, 2020
  • NCUMC — Second Diploma, 2019