I'm a deep learning engineer and AI researcher driven by a simple belief:
"Technology should lift burdens, not just shift them."
I build intelligent systems that empower industries and research while staying rooted in human impact.
Industrial AI
Turning industrial data into production-ready AI. At Merck I raised first-time-right rates in photoresist blending and rebuilt a pigment-line pipeline that won the 2024 Surface Gernsheim Efficiency Award.
CV & Robotics
Bridging AI with the physical world. I design multimodal visual representations for robotic manipulation and medical imaging, empowering machines to solve complex real-world challenges.
Human-Centered AI
I build intelligent systems that people actually want to use. By combining deep learning with intuitive frontend interfaces, I make complex algorithms accessible to non-technical domain experts.
AI & Vision
Deep Learning
PyTorch, Imitation Learning
CV & Robotics
Embodied AI, DINO, 6D Pose
Generative AI
Diffusion Models, VAE, ComfyUI
Development
Frontend
Vue.js, Next.js, Tailwind CSS
Backend
FastAPI, Python, RESTful APIs
HPC & DevOps
SLURM, Docker, AWS, Linux
Data Science
Processing & ML
Pandas, XGBoost, Scikit-learn
Interactive Viz
Streamlit, Plotly, BertViz
Other Skills
NLP & XAI
LangChain, Hugging Face, XAI
Research Practice
Ablation design, failure taxonomies, reproducibility
Growing up witnessing medical gaps and workers shoulder society's hardest tasks, I design AI that reduces invisible labor.
Whether optimizing industrial production or tackling challenges with code, I measure success by how many people breathe easier.
Seeking immediate opportunities in:
PhD Positions
Self-supervised and object-centric representations, robustness under shift
Industrial Research
Robot perception and manipulation, representation learning under shift
Vision & Medical AI
Detection and segmentation under scarce labels and domain shift
My Projects
Building solutions that make a difference
Self-Supervised Learning of a Visual Object-Centric Representation for Robotic Manipulation
TU Darmstadt (IAS Lab) × École Centrale de Lyon (LIRIS) • October 2024 - April 2026 | Prof. Jan Peters, Prof. Liming Chen, Alexandre Chapin, Alap Kshirsagar
M.Sc. thesis, TU Darmstadt — submitted 23 March 2026, defended 17 April 2026, graded 1.3 (sehr gut). Investigated how frozen, self-supervised visual representations can be integrated into continuous robotic manipulation policies under standard hardware constraints (single RTX 2080 Ti).
RARE26 — Early Barrett's Neoplasia Detection Challenge
EndoVis · MICCAI 2026 — solo entrant • July – September 2026 | Solo entrant · code and technical report public under MIT
A solo challenge entry in medical imaging: a frozen backbone, a 4096-parameter read-out head, and 1835 pre-registered cross-center comparisons — 77% of which resolved nothing, and are published anyway.
Repose via Image Translation: 6D Pose & Synthetic Data
Fraunhofer IGD / Graphisch-Interaktive Systeme (GRIS Lab), TU Darmstadt • November 2022 - August 2023 | Thomas Pöllabauer
Architected an automated synthetic data generation pipeline on a SLURM cluster and evaluated generative image translation models for industrial 6D pose estimation.
Semantic Segmentation Diffusion Models
Graphisch-Interaktive Systeme (GRIS Lab) • March 2023 - October 2023 | Yannik Frisch
Improved semantic segmentation accuracy in medical images using diffusion models to generate high-quality synthetic data for augmenting.
SQuARE - Explainability Methods for QA
Ubiquitous Knowledge Processing Lab (UKP Lab) • October 2022 - February 2023 | Haritz Puerto
Improved explainability of AI models in NLP by extending explainability methods on UKP-SQuARE platform for QA pipelines.
Endoscopy OOD Augmentations
Medical & Environmental Computing (MEC-Lab) • December 2022 - February 2023 | Henry Krumb
Enhanced Out-of-Distribution (OOD) detection for endoscopy images by exploring different OOD augmentation methods and image attribute shifts.
E-ScootAR: Unimodal Warnings for E-Scooter Riders
Telecooperation Lab (TK) • April 2021 - July 2021 | Andrii Matviienko
Evaluated technological safety measures for e-scooter riders using augmented reality notifications, vibrotactile feedback, and auditory signals.
Molecular Docking Study of TBBPA in Liver Cancer
Institutes of Biomedical Sciences, Shanxi University • March 2018 - November 2018 | Zhuoyu Li
Investigated the regulation of calcium-mediated lysosomal exocytosis and its impact on cell invasion and migration in liver cancer.
Publications
Peer-reviewed work and the thesis behind it
- First authorAccepted
More Structure, Not More Capacity: Object-Centric Representations for Visuomotor Imitation Learning
Yi Li, Alexandre Chapin, Liming Chen, Jan Peters, Alap Kshirsagar
RSS 2026 Workshop From Perception to Action · poster · non-archival
arXiv: 2607.09825
- Co-author
E-ScootAR: Exploring Unimodal Warnings for E-Scooter Riders in Augmented Reality
ACM CHI 2022 Extended Abstracts · DOI 10.1145/3491101.3519831
- Co-author
TBBPA regulates calcium-mediated lysosomal exocytosis and thereby promotes invasion and migration in hepatocellular carcinoma
Contributed molecular docking (AutoDock / PyMOL) and statistical analysis
Ecotoxicology and Environmental Safety, 2020 · DOI 10.1016/j.ecoenv.2020.110255
Self-Supervised Learning of a Visual Object-Centric Representation for Robotic Manipulation
M.Sc. thesis, TU Darmstadt, 2026 · Reviewers: Prof. Jan Peters, Prof. Liming Chen · Supervisors: Alap Kshirsagar, Alexandre Chapin
Grade 1.3 (sehr gut)
Tailored CV on request — [email protected]
My Journey
Code with Compassion, Automation for Liberation
Work Experience
Working Student, Data Sciences
Merck KGaA, Darmstadt, Germany
Darmstadt, Hessen, Germany
Oct 2023 - Sep 2025
Applied machine learning and full-stack engineering to semiconductor and chemical manufacturing across four projects — earning two company awards and a reference letter signed by two Directors.
🏆 Surface Gernsheim Award 2024 (Efficiency)
🏆 Spot Award 2025 (MyImpact@Merck)
Core Projects: Photoresist formulation prediction (ML pipeline + Streamlit) · Pigment production optimization (FRED 2.0) · Pharmaceutical parameter regression · Full-stack lab reservation system (Vue.js + FastAPI + AWS).
Work spanned the full delivery chain: legacy SQL/Python data extraction, ML model development, interactive web application deployment, and cross-functional collaboration with chemical engineers.
Working Student, AI/VR
NMY Mixed Reality Studio
Frankfurt am Main, Hessen, Germany
Jun 2023 - Sep 2023
Built a retrieval-augmented voice assistant prototype (LLaMA 2 + LangChain + ChromaDB, Bark TTS) over an internal knowledge base. A hardware feasibility assessment showed local multi-user inference exceeded the available compute; I handed that scalability finding to the team.
Co-founder and Product Manager
Hangzhou BseTech Co., Ltd.
Hangzhou, Zhejiang, China
Oct 2015 - Jan 2016
As the co-founder and product manager of a wellness industry startup, I explored every step of the process from company registration to launching the initial apps.
Education
M.Sc. Computer Science
Technische Universität Darmstadt
Darmstadt, Hessen, Germany
Apr 2019 - Apr 2026
Focus on AI and deep learning, specializing in computer vision and ML systems. Minor in Entrepreneurship & Innovation. Master's thesis on object-centric visual representations for robotic manipulation, graded 1.3 (sehr gut), defended April 2026.
Core Courses:
Key Seminars:
- Deep Learning for Medical Imaging (02/2023) - Approval and Regulation of AI Medical Devices
- Deep Generative Models (07/2022) - OOD Detection with Outlier Synthesis
- Advanced CV & ML (04/2022) - Semantic Scene Segmentation with Context
- Deep Learning & Digital Humanities (01/2022) - Lexical Semantic Change
B.Sc. Computer Science
Chongqing University of Post and Telecommunications
Chongqing, China
Sep 2012 - Jul 2016
Sino-Canadian joint program. Computer science fundamentals with early exploration of machine learning. Ranked 1st in the department in freshman year; awarded an RMB 8,000 scholarship for academic excellence.
Campus Activities:
- Minister of Publicity Department, Students' Science and Technology Union (2013-2014)
- Trumpeter in the University's Orchestra, College Students Art Troupe (2012-2015)
- Product Designer for KuaiShou APP, Shortlisted in Chongqing Youth App Development Competition (2014)
- Campus Representative for Meituan (2014)
Technical Arsenal
Crafting solutions with precision tools and innovative approaches