Tigran Galstyan

Tigran Galstyan

πŸŽ“ PhD Candidate in Computer Science | 🧠 Machine Learning Researcher

πŸ“ 3/11 Avan-Ariuj, Yerevan, Armenia 0022
πŸ“± +374 (91) 00-45-49 | πŸ“§ galstyantik@gmail.com
πŸ”— LinkedIn | GitHub | Google Scholar


πŸŽ“ Education

PhD in Computer Science
Institute for Informatics and Automation Problems of NAS RA
πŸ“… 2020–2024 | Adviser: A. S. Dalalyan
πŸ“œ Thesis: β€œStatistical and Computational Complexity of the Feature Matching Map Detection Problem”

Master of Science in Discrete Mathematics & Theoretical Informatics
Bachelor of Science in Computer Science & Applied Mathematics
Yerevan State University
πŸ“… 2014–2020 | πŸ… GPA: 19/20 (Both Degrees)

Physical and Mathematical Specialized School
πŸ“… Graduated 2014 | πŸ₯‡ Valedictorian (Top of 140 students) | GPA: 9/10


πŸ’Ό Professional Experience

YerevanN Research Lab

🧠 Machine Learning Researcher | πŸ“… July 2016 – Present

πŸ“ Yerevan, Armenia

  • Led Aerial Vision-and-Language Navigation for UAVs using multimodal models (LLaVA, Chameleon)
  • Developed statistical methods for feature matching map recovery (PaperCode)
  • Pioneered work on domain generalization failures (Code)
  • Built BioreIEx biological relation extraction system (ACL Paper)
  • Created LSTM-based transliteration models (Blog)

American University of Armenia

πŸ‘¨πŸ« Adjunct Professor | πŸ“… Sep–Dec 2021

  • Taught DS115 Data Structures/Algorithms to 130 students
  • Adapted Stanford’s CS 161 curriculum for Python-based competitive programming

🌟 Google Internships

X, the Moonshot Factory

πŸ€– Software Engineering Intern | πŸ“… Jun–Aug 2019 | πŸ‡©πŸ‡ͺ Munich

  • Bridged simulation-to-reality gaps for The Everyday Robot project

Google NYC

πŸ“Ή Software Engineering Intern | πŸ“… Aug–Oct 2018 | πŸ‡ΊπŸ‡Έ New York

  • Improved YouTube captioning with contextual biassing
  • Designed named entity transcription evaluation metrics

Google Switzerland

✍️ Research Intern | πŸ“… Jul–Sep 2017 | πŸ‡¨πŸ‡­ Zurich

  • Enhanced handwriting recognition via adversarial training

πŸ“š Publications

  1. Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution
    Vardanyan E., Hunanyan S., Galstyan T., et al. (2024). ICML
  2. Matching Map Recovery with an Unknown Number of Outliers
    Minasyan A., Galstyan T., et al. (2023). AISTATS
  3. Failure Modes of Domain Generalization Algorithms
    Galstyan T., Harutyunyan H., et al. (2022). CVPR
  4. Biorelex 1.0: Biological Relation Extraction Benchmark
    Khachatrian H., Galstyan T., et al. (2019). ACL BioNLP Workshop

(10+ total publications in ICML, CVPR, EJS, and more)


πŸ† Honors & Awards

  • πŸ₯‰ Bronze Medal, International Olympiad in Informatics (IOI 2014)
  • πŸ₯‰ Bronze Medal, Zhautikov Olympiad in Informatics (2014)
  • πŸ… Finalist, Google Hashcode (2017, 2018)

πŸ›  Skills

| Category | Details | |β€”β€”β€”β€”β€”β€”β€”|————————————————–| | Programming | Python (NumPy, PyTorch), LLMs, Neural Networks | | Machine Learning| Domain Generalization, Vision-Language Models | | Languages | Armenian (Native), English, Russian (Fluent) |


πŸ“œ References

  1. Arnak Dalalyan
    Director at CREST, France | PhD Co-Advisor
  2. Aram Galstyan
    Distinguished Principal Scientist, USC ISI | Research Mentor