About

InterDigital

Co-leading the AI codec team in London following InterDigital's acquisition of Deep Render. Advancing next-generation neural video compression technologies, combining decades of video research with cutting-edge AI to drive the future of how video content is delivered.

Deep Render

Technical lead and manager of 15 engineers responsible for productization of research models and internal hardware/software infrastructure. Built unified model inference library for major AI accelerators (Nvidia, Apple, Qualcomm, Intel) and CI/CD system reducing model port times from weeks to days. Spearheading the world's first integration of a neural codec into FFmpeg and VLC achieving realtime video encode and decode on consumer devices.

Deep Render

Led model quantization and pruning research achieving 2x improvements in memory footprint and runtime while retaining compression performance. Delivered client-facing applications and built model evaluation tools using Python and web stack (Vue, Node).

Deep Render

Made foundational research contributions to video compression models with >10% efficiency gains over state-of-the-art. Lead maintainer of core model codebase (PyTorch, C++, CUDA) with 60k+ lines of contributions, focusing on generative CV models and optical flow.

Sledilnik COVID-19 Tracker

Sledilnik is an open source community engaged in a comprehensive effort to track COVID-19 cases in Slovenia. Contributed to data visualizations using the backend written in F# and created visualizations using the Highcharts API.

Method and Data Processing System for Lossy Image or Video Encoding. Motion translations with flow-based processes for AI video compression.

Method and Data Processing System for Lossy Image or Video Encoding. Motion transformation handling in AI compression processes.

Covering advanced neural compression techniques, optical flow, and video encoding optimization methods.

SciPost Physics

With M. van Beest, S. Schafer-Nameki, J. Sparks

Journal of High Energy Physics

With A. Braun, M. Hubner, S. Schafer-Nameki

University of Oxford

Focused on the intersection of geometry and string theory, exploring higher-dimensional geometric objects that appear in string theory and how their properties manifest in associated physical theories.

Imperial College London

Advanced studies in pure mathematics, building the foundation for doctoral research.

University of Ljubljana

Undergraduate studies in mathematics with a focus on theoretical foundations.

Probability & Statistics

Highly proficient in mathematics, probability theory, and statistics, with a particular interest in generative density models.

Deep Learning

Highly proficient in applied generative AI research both in computer vision and applications of LLMs for internal business infrastructure projects.

Programming

Python (advanced) - 10+ years in software development, data analysis, and AI research. Expert in PyTorch with deep knowledge of model compilation stack (FX, Inductor, Dynamo, Triton).

C++ (working knowledge) - Contributions to model inference libraries and entropy coders.

I'm always interested in discussing machine learning, mathematics, and technology.

• Email sebastjancizel@gmail.com
• LinkedIn /in/scizel
• GitHub @sebastjancizel
• Resume Download PDF