Tigran Galstyan

Tigran Galstyan

PostDoctoral Research @ L2S - CentraleSupélec

Tigran Galstyan

Tigran Galstyan

Education

Institute for Informatics and Automation Problems of NAS RA
PhD Thesis Defence on “Statistical and Computational Complexity of the Feature Matching Map Detection Problem”
Advised by A. S. Dalalyan
May 2024

Yerevan State University, Yerevan, Armenia
Master of Science in Discrete Mathematics and Theoretical Informatics
GPA 19/20
June 2020

Yerevan State University, Yerevan, Armenia
Bachelor of Science in Computer Science and Applied Mathematics
GPA 19/20
June 2018

Physical and Mathematical Specialized School, Yerevan, Armenia
Valedictorian in class of 140 students
GPA 9/10
June 2014

Experience

YerevaNN Research Lab

Machine Learning Researcher
July 2016 - Present
Yerevan, Armenia

  • Currently working on Aerial Vision-and-Language Navigation for UAVs
  • Ongoing work is based on multimodal models (e.g. LLAVA, Chameleon) and exploration of their capabilities for Vision-Language Navigation problem
  • Worked on PhD thesis on statistical properties of the problem of feature matching map detection problem
    • Main paper: https://proceedings.mlr.press/v206/minasyan23a/minasyan23a.pdf
    • Code: https://github.com/TigranGalstyan/Matching-Map-Recovery-with-Outliers
  • Worked on problems of domain generalization and invariant representation learning
    • Papers:
      • Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution
      • Failure Modes of Domain Generalization Algorithms
    • Code: https://github.com/YerevaNN/dom-gen-failure-modes
  • Part of the BioRelEx project on biological relation extraction
    • Paper: https://www.aclweb.org/anthology/W19-5019/
    • Code: https://github.com/YerevaNN/BioRelEx
  • Implemented and compared sequence to sequence techniques for transliteration (from romanized Armenian to Armenian script)
    • Blogpost: https://yerevann.github.io/2016/09/09/automatic-transliteration-with-lstm/
    • Code: https://github.com/YerevaNN/translit-rnn

American University of Armenia

Adjunct Professor
September 2021 - December 2021
Yerevan, Armenia

  • Taught DS115 Data Structures/Algorithms in Data Science sophomore course to class of 130 students
  • Course based on Stanford University’s CS 161 Design and Analysis of Algorithms by Karey Shi
  • Covered foundational ideas of theory of data structures and algorithms with practical coding in Python for competition-style programming problems (e.g. HackerRank, Codeforces)

Google’s X, the Moonshot Factory

Software Engineering Intern
June 2019 - August 2019
Munich, Germany

  • Worked on bridging the gap between simulation and real-world model performance on The Everyday Robot project

Google NYC

Software Engineering Intern
August 2018 - October 2018
New York, United States

  • Worked on a project to utilize contextual biassing in the YouTube captioning system
  • Developed and implemented various named entity transcription evaluation techniques
  • Conducted experiments and measured the effects of biassing on YouTube captioning

Google Switzerland

Research and Machine Intelligence Intern
July 2017 - September 2017
Zurich, Switzerland

  • Worked on Google’s handwriting recognition
  • Implemented adversarial training techniques and made them accessible across the team

Teamable Software

Machine Learning Intern
July 2016 - September 2016
Yerevan, Armenia

  • Presented papers of state-of-the-art algorithms for learning word embeddings
  • Implemented and used machine learning algorithms such as Brown clustering, Word2Vec, FastText

Publications

  • Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution. Vardanyan E., Hunanyan S., Galstyan T., Minasyan A., Dalalyan S. A. (2024). Proceedings of the 41st International Conference on Machine Learning, PMLR 235: 49203-49225
  • Matching Map Recovery with an Unknown Number of Outliers. Minasyan, A., Galstyan, T., Hunanyan, S., & Dalalyan, A. (2023). In International Conference on Artificial Intelligence and Statistics, PMLR, 891-906.
  • Optimality of the Least Sum of Logarithms in the Problem of Matching Map Recovery in the Presence of Noise and Outliers. Galstyan, T. & Minasyan, A. (2023). Armenian Journal of Mathematics, 15(5), 1-9.
  • Comparison of data matching methods on biomedical datasets. Galstyan T. (2022), Vestnik of Russian-Armenian University, 1(2):46–58.
  • Failure Modes of Domain Generalization Algorithms. Galstyan, T., Harutyunyan, H., Khachatrian, H., Steeg, G. V., & Galstyan, A. (2022). In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 19077-19086.
  • Optimal detection of the feature matching map in presence of noise and outliers. Galstyan, T., Minasyan, A., & Dalalyan, A. (2022). Electronic Journal of Statistics, 16(2):5720–5750.
  • Robust Classification under Class-Dependent Domain Shift. Galstyan, T., Khachatrian, H., Steeg, G. V., & Galstyan, A. (2020). ICML Workshop on Uncertainty & Robustness in Deep Learning.
  • On the Tradeoff Between Accuracy and Fairness in Representation Learning. Galstyan, T. & Khachatrian, H. (2020), Vestnik of Russian-Armenian University, 1:35-46.
  • On Random Weighted Sum of Positive Semi-Definite Matrices. Galstyan, T. & Minasyan, A. (2020). Proceedings of the YSU A: Physical and Mathematical Sciences, 54(2 (252)), 96-100.
  • Biorelex 1.0: Biological relation extraction benchmark. Khachatrian, H., Nersisyan, L., Hambardzumyan, K., Galstyan, T., Hakobyan, A., Arakelyan, A & Galstyan, A. (2019). In Proceedings of the 18th BioNLP Workshop and Shared Task, 176-190.

Honors and Achievements

  • Bronze Medal, International Olympiad in Informatics (IOI), 2014, Taipei, Taiwan
  • Bronze Medal, International Zhautikov Olympiad in Informatics, 2014, Almaty, Kazakhstan
  • Finalist, Google Hashcode 2018, Dublin, Ireland
  • Finalist, Google Hashcode 2017, Paris, France

Leadership and Activities

Armenian National Olympiad in Informatics

Jury Member/Tutor
October 2014 - May 2016
Yerevan, Armenia

  • Trained high school students for national and international programming competitions
  • Coordinated the national informatics olympiad with over 100 students from various regions of Armenia
  • Prepared tasks and tests for the national competition and team selection contest for IOI

Skills

  • Programming: Python (NumPy, PyTorch), Machine Learning, Neural Networks, Large Language Models
  • Languages: Armenian (Native), English, Russian (Fluent)

References

  • Arnak Dalalyan, Director at CREST, FRANCE
    • PhD co-advisor, worked on my PhD thesis, co-authored several papers in major publications
  • Aram Galstyan, Research Adjunct Professor, Research Director, Distinguished Principal Scientist at USC ISI
    • Supervised work on Biological relation extraction project, published a paper in ACL BioNLP workshop