About Us
Established in 2016, at the YerevaNN Scientific-Educational Foundation (YerevaNN), a non-profit computer science research center, we are enabling an ML world-class research in Machine Learning (ML) in Armenia.
The mission of YerevaNN is to stimulate scientific careers in computer science and to facilitate the growth of the next generation of computer science professors in Armenia. YerevaNN succeeds with its mission by building a research-friendly environment for early-career computer scientists and setting up collaborations with the world's leading labs.
Excellent Research
While the high-tech industry is relatively well developed in Armenia, there are very few high-quality research labs that are accessible for students who study computer science.
At YerevaNN university students get involved in world-class scientific research since their undergraduate studies, collaborate with well-known scholars from various countries, and publish papers in leading journals and conferences.
YerevaNN makes it possible for students to choose academic careers, or to become the next generation of AI leaders in the industry while staying in Armenia.
Compute Infrastructure
In 2024, we significantly enhanced our computing capabilities with the addition of 8 NVIDIA H100 GPUs with the funding of the RA Science Committee. This infrastructure, hosted at Yerevan State University (YSU), serves as our primary computing platform. The current resources allow us to: - Train ~1B parameter models like Llama 3.2 1B on 30B+ tokens in a day. - Conduct fast experimentation on hyperparameters at that scale and optimize model performance. - 3B parameter training is also within reach in a reasonable time-scale. - It is also possible to train up to 10B models in under two weeks. The availability of high-performance computing has transformed our research capabilities by accelerating the experimentation speed, optimizing resource allocation for development and debugging. It created a need for additional development systems for daily operations.
A major infrastructure expansion is underway with the procurement of 64 additional H100 NVIDIA GPUs through government funding. YSU will house and manage this expanded infrastructure. GPUs are expected to arrive in December 2024 and fully operational by September 2025.
Research Environment
A strong research environment is key to driving innovation and fostering professional growth. Engaging with the global research community through conferences, workshops, and summer schools not only helps exchange ideas but also strengthens collaborations and inspires new directions. Our team actively participates in international and regional events, showcasing their work, learning from others, and contributing to advancing the field.
Featured in Media
DataFest Yerevan
Alla Barseghyan presents our research on Less is More Data Specialization for Various Modalities at DataFest Yerevan
Rock Talk
[Armenian] Կփոխարինի՞ արդյոք AI-ը ծրագրավորողներին. AI-ի ապագան և սահմանափակումները
DataFest Yerevan
Hakob Tamazyan: Benchmarking Robustness of Foundation Models for Remote Sensing
Network Nation
Hrant Khachatrian talks about YerevaNN's current and future plans and motivation to do research.
EVN Report
Hrant Khachatrian discusses the lab’s various research directions and their strategic importance for Armenia
YandexHall
ML OpenTalk: Speeding up training with FP8 and Triton Vlad Savinov, Team Lead, YandexGPT pretraining, presents the key principles behind speeding up model training.
DataFest Yerevan
Khoren Petrosyan presents our research on "From 0 02% Samples to Full Radio Maps: Physics‑Informed U‑Net for Indoor Path" at DataFest Yerevan
Rearrange
Hrant Khachatrian sat with Rearrange's Narek Amirkhanian to talk about the AI new models, the price of technological developments and more.
Lratvakan Radio
The first Large Language Models (LLM) Summer School has launched in Yerevan.
DataFest Yerevan
Tigran Fahradyan: Small Molecule Optimization with Large Language Models
tech.news.am
DeepSeek–ի հաջողությունը, արհեստական բանականության հայկական մշակումները. Հրանտ Խաչատրյան
Bun TV
Our collaborator Vahan Huroyan recently sat down Bun TV and had an engaging conversation about his path to becoming a scientist.
DataFest Yerevan
Hasmik Mnatsakanyan: Exploring the Recall of Language Models: Case Study on Molecules