21 top students from all around the globe are now in Sofia for this year’s INSAIT Summer Research Fellowship Program 2024! Amongst them are international olympiad medalists, startup founders, gen AI hackathon winners and others. The research topics they work on at INSAIT span large language models, 3D understanding, safety of generative AI, security of machine learning, quantum computation, and more. Learn more about these amazing students and their projects below:
21 top students from all around the globe are now in Sofia for this year’s INSAIT Summer Research Fellowship Program 2024
Sanjana Garai – at INSAIT, she is developing a visual chatbot that uses fine-tuned Visual Language models for various tasks. She created her own startup, aimed to boost sustainability in fashion and got selected for the semi-finals of the HULT Prize 2023, being one of the top 70 startups across the 2000+ teams from the Asia-Pacific.
Timofei Fedoseev – аt INSAIT, he is enhancing LLMs’ reasoning abilities for solving mathematics tasks. He is a gold medalist at the International Olympiad in Informatics (IOI) in 2021 and an absolute winner at the All-Russian Olympiad in Informatics in 2021. He is also a gold medalist at the ICPC World Finals 2023.
Aleksandar Yanev – his project at INSAIT is about generating videos based on text. It aims to solve the task of making videos look physically realistic. Aleksandar is a medalist at the Romanian Masters of Informatics and a silver medalist at the 2024 Mathematics Olympiad for University Students.
Koichi Namekata – at INSAIT he is investigating interactive 3D scene representations. He won bronze and silver medals at the International Olympiad in Informatics (IOI).
Pramish Paudel – his summer project aims to improve DALL-E’s precision in editing images from text descriptions, allowing specific modifications without altering the entire image. He won the Hult Regionals Summit, securing a spot among the top 16 teams selected from among more than 30,000 teams worldwide. Co-author of an ECCV’24 paper.
Joanna Cheng – at INSAIT she is working on adding semantic information to 3D Gaussians, which can enhance downstream applications such as 3D segmentations, editing, etc. She is a top B.Sc. student in Computer Science, Applied Mathematics and Statistics at Johns Hopkins University, USA.
Emma Anastassova – at INSAIT she is working on a new framework for evaluating large language models (LLMs). She is a Bocconi AI and Neuroscience research representative at World AI Cannes Festival 2024.
Marwa El Kamil – at INSAIT, she is working on LLM tokenization to better handle languages with different alphabets, improving fine-tuning and inference speed while maintaining accuracy. Her team won the GenAI Hackathon-MoroccoAI in 2023 for creating an LLM-powered chatbot for Moroccan farmers.
Thien-Phuc Tran – his project at INSAIT focuses on photorealistic reconstruction and rendering of 3D articulated robots/agents onto robotics simulators, and achieving full manipulation of the rendered models in real time. As an undergraduate researcher his latest paper was published at the MediaEval 2023 workshop.
Emircan Gundogdu – at INSAIT, he is researching how training patterns in neural networks can enhance learning and generalisation. One of his achievements is getting a full scholarship for MSc at Imperial College London – only given to one Turkish student each year.
Dimitar Chakarov – at INSAIT, he and his team use a game-theoretic approach to study how differing client objectives affect federated learning outcomes and explore mechanisms to disincentivize bad behaviours. He is a winner of the 2019 John Atanasoff Award for Research Debut in Computer Science and a recent Princeton graduate.
Negin Rahimi-Yazdi – her project focuses on performative mean estimation in machine learning theory. It explores how our predictions of data statistics can influence the phenomena being predicted. Ranked 1st among 82 computer science students at Amirkabir Technical University, Iran
Csaba Dékány – at INSAIT he is working on making large language models resilient to attacks, so to prevent them from answering unethical or illegal questions. Medalist from the International Physics Olympiad (IPhO) and the European Physics Olympiad (EuPhO) in 2022.
AmirHossein Razlighi – his project at INSAIT is focused on reconstructing 3D scenes from a set of 2D images captured from different viewpoints of that scene. Ranked top 100 amongst more than 150,000 students in the entrance exam at Sharif Technical University, Iran.
Mahdi Saber – at INSAIT, he is designing a verified compiler for quantum programing. Mahdi was a gold medalist in the Zanjan Mathematics Competition and placed 6th in the National Mathematics and Physics Examination in Iran among more than 50,000 participants.
Darin Krumov – at INSAIT, he focuses on language embeddings used for models such as BgGPT. Darin was a medalist in the 2017 to 2022 national high school math competitions.
Naser Kazemi – at INSAIT, he is developing a model for interactive video generation that creates smooth, realistic video sequences from an initial frame or sketch based on described actions. He was a gold medalist at the European Mathematical Cup in 2018 and ranked 5th in the national university entrance exam in 2020.
Milan Stanković – at INSAIT, he is exploring the security of large language models. He won a silver medal 3 times (2019, 2021, 2022) at the National Physics competitions in Serbia, consistently ranking among the top 10 students in the country.
Siavash Rahimi Shateranloo – at INSAIT, he is designing novel leakage attacks for federated learning. He is an Iranian Mathematics Olympiad gold medalist and a bronze medalist at the Romanian Masters of Mathematics.
Yitong Ma – his project at INSAIT aims to optimise UMAP by improving its speed, reducing memory usage, and exploring incremental update methods for efficiency compared to reloading from scratch. He has been awarded the First-Class Scholarship at Zhejiang University for 3 consecutive years and received the 2nd prize at the National College Student Mathematics Competition of China in 2022.
Anikesh Parashar – His research focuses at INSAIT is on uncomputation of unwanted intermediate values in quantum computing. He is ranked 24th out of around 50,000 in KVPY 2020 (Indian National Science Aptitude Examination) and 5th in ISIAT 2021 (Indian Statistics Entrance Examination).
Learn more about them here