📢 Note: Are you an Engineer, Researcher, or Entrepreneur looking for feedback on your project? I’m here to help. I love to play around with new libraries, gadgets, models, and APIs and happy to provide honest and constructive feedback. Feel free to ping me at dsoselia{at}umd.edu. Excited to see what you’re working on!

Hi there! My name is Davit and I am a PhD Candidate at the University of Maryland, College Park. I am advised by Dr. Amitabh Varshney and collaborate closely with Dr. Tianyi Zhou. I have a passion for solving problems with machine learning and am always looking for ways to better understand how models work and how we can improve them. I am excited to embark on this new journey and hope that it will help me progress in both of these areas.

I enjoy all areas of machine learning, but my current interests include VLMs, large models, and vision-language alignment. I am also interested in exploring the intersection of machine learning with other fields such as document analysis, code generation, and software security.

Before starting my PhD, I was based in Sweden and have worked as a machine learning engineer and before that as a software engineer for over 3 years. I have led projects in machine learning and software development, and have worked with clients in the US and the EU.

Thank you for reading through my page. If you have any questions or just want to get in touch, please don’t hesitate to reach out to me via my email or social media. I am always happy to chat with other researchers and tech enthusiasts and share ideas and experiences.

What’s New?

• Our work has been accepted at BMVC2024. Looking forward to connecting with the fellow reserchers in Glasgow.

• Had a great summer at Adobe Research, stay tuned for some exciting updates!

• Our work “Odin: Disentangled reward mitigates hacking in rlhf” accepted at ICML 2024.

• 2023-07-28: Our work “Estimation of Joint Torque by EMG-Driven Neuromusculoskeletal Models and LSTM Networks” accepted at IEEE Transactions on Neural Systems and Rehabilitation Engineering.

• 2023-06-06: Our work “Vision-Code Transformer for Screenshot-to-HTML/CSS Generation” Accepted at T4V @ CVPR 2023.

• 2022-09-05: I moved to College Park, MD to begin studying Computer Science at the University of Maryland. I am excited to return to an academic setting and have the opportunity to dive deeper into topics that I have been interested in for a while. I’m also looking forward to exploring a new city and meeting new people.This move marks the three-year anniversary of my arrival in Sweden, and I am grateful for the time I have had here and the amazing people I have met, worked with, and studied with. Some of the highlights of my time in Sweden include:

• Collaborating with Professor Hossein Azizpour, where I had the opportunity to focus on exciting deep-learning research that I hope to continue in the future.

• Collaborating with Professor Elena Gutierrez Farewik at the Movability Lab at KTH to try to better approximate human movement with machine learning.

• Google TVC assignment, working on some really cool stuff, that unfortunately I can talk little about.

• Developing FinSentim, a tool for deep learning-based financial sentiment analysis, with my friends and colleagues.

• Working on Chromosome and genetic data analysis using ML together with amazing team at Arkus AI.