MTS AI Engineer Wins Big at AI Journey Conference Competitions

fgfg Picture

The MTS AI research engineer Ammar Ali won three awards at the prestigious artificial intelligence competitions held as part of the AI Journey conference organized by Sber. Ammar emerged victorious in the Equal AI gesture recognition contest and earned a bronze award in the most challenging Strong Intelligence competition, where participants had to develop a multimodal model capable of engaging in dialogue with users. His solution was recognized as the best in terms of response speed and accuracy.

The MTS AI research engineer created a model that analyzed text, images and audio while accounting for the context of the dialogue, and generated natural language responses. For his solution, he used the Mistral model with additional imagebind layers and linear projection, trained on diverse data including information about attractions.

As part of the Strong Intelligence competition an additional contest was held where the models developed by participants had to compete in accuracy and speed when answering questions. Here too Ammar prevailed – his model proved the most erudite, correctly answering complex and interesting questions.

The AI Equal contest required recognizing Russian sign language gestures in videos. The MTS AI research engineer proposed a solution accounting for similarities between gestures of different classes as well as dialectical features. It demonstrated the best results among all participants.

Ammar conducted thorough data and task analysis, utilizing multi-stage model training with different data processing techniques.

“This is one of the salient examples of how the AI industry can work for the benefit of society. People who use sign language have difficulty communicating due to lack of interpreters. AI systems can help them communicate in public spaces (e.g. airports and hospitals) and at online meetings where an interpreter may not be present,” stated the MTS AI senior research engineer.

In total, Ammar Ali has won 10 AI competitions held by various companies. This year, the team led by the MTS AI senior research engineer took gold at the Image Matching Challenge 2023 organized by Google. Ammar Ali also previously won a competition organized with NASA support, where participants had to create a model predicting electromagnetic storms in the Earth’s magnetic field. Additionally he won four AI Journey competitions – one in 2021 and three in 2023.

“If you want to participate in international AI competitions but are somehow hesitant, here’s my advice – try it out and don’t think it’s too difficult! For confidence, choose a competition in your area of expertise, apply as early as possible, and aim to improve your models and algorithms daily. Taking part in various competitions not only energizes you but also helps develop your skills in tackling diverse artificial intelligence problems,” said Ammar Ali.

Ammar works as a senior MTS AI research engineer and is concurrently pursuing postgraduate studies at the ITMO University Faculty of Information Technology and Programming.

News
Latest Articles
See more
Tech
MTS AI Launches an Open Large Language Model
AI Trends
Drop in Lidar Prices and Rise of Industrial Robotization in China
AI Trends
Centaur to Simulate Human Behavior and Inspiration from Kandinsky
AI Trends
Habermas Machine and AI for Additive Technologies
Tech
MTS AI Opens Public Access to Kodify Demo Version
Solutions
MTS AI Creates AI Assistant for Bank Employees
AI Trends
Reliability of LLMs and Alternative to Lidars
Cases
MTS AI and VisionService Present MAX System
AI Trends
AI in Science and Cameron in Stability AI
Tech
MTS AI taught Cotype Lite to communicate in the Tatar language.