Martin Juan José Bucher
PhD Student at Stanford University
Office: Y2E2 271B
Gradient Spaces Lab
Stanford, California (US)
Twitter // Scholar // LinkedIn // GitHub
About
Hi, I'm Martin, a second-year PhD Student at Stanford University. My research interests lie at the intersection of Deep Generative Modeling, Representation Learning, and Computer Vision — primarily centered on leveraging Generative Models and Scene Synthesis for a Circular Built Environment. This endeavor is shaping a new frontier that I envision as Computational Circularity.
News
- 08/21/24 — I have been selected into the Stanford HAI Graduate Fellowship Program for the 2024/2025 cohort!
- 06/25/24 — We finally published a paper on arXiv on fine-tuned encoder-style LLMs vs. decoder-style LLMs for text classification, which has been sitting in Overleaf for far too long (and has nothing to do with my PhD research). Paper
- 10/20/23 — My first paper as first author has been accepted in Automation in Construction! Topic: Performance-Based Generative Design using CVAEs. Paper
- 09/19/23 — First day at Stanford University. Let the PhD journey begin...
- 09/14/23 — Last week in Europe before packing up! We presented two Posters at the 2nd Future of Construction Symposium at TU Munich (TUM). Some photos here
- 05/04/23 — I will be joining Stanford in September as a PhD Student in the newly formed Gradient Spaces Lab under Prof. Iro Armeni. Off across the Atlantic ocean!
- 01/01/23 — Started as a Research Assistant at ETH Zurich in the CEA Lab to drive the Mapping Ukraine project under the supervision of Prof. Catherine De Wolf.
- 04/18/22 — Finished my Master's Thesis at ETH Zurich in collaboration with Design++. My research bridges the fields of Generative Design and Deep Generative Modeling. Paper is under review and I can hopefully share some results soon!
Publications
-
Fine-Tuned 'Small' LLMs (Still) Significantly Outperform Zero-Shot Generative AI Models in Text Classification
-
Performance-Based Generative Design for Parametric Modeling of Engineering Structures Using Deep Conditional Generative Models
-
Towards a ‘Resource Cadastre’ for a Circular Economy– Urban-Scale Building Material Detection Using Street View Imagery and Computer Vision
-
GENEVIZ: A Visual Tool for the Construction and Blockchain-Based Validation of Service Function Chaining (SFC) Packages
Longer Bio
Martin Juan José Bucher is a PhD Student at Stanford University and part of the Gradient Spaces Lab, advised by Prof. Iro Armeni. He is currently based in Stanford, California (US). Before crossing the Atlantic ocean to start this journey, he worked as a Research Assistant (2023) at ETH Zurich in the Circular Engineering for Architecture (CEA) lab under Prof. Catherine De Wolf, fusing ideas from both Computer Vision and Circular Engineering. Martin graduated with an MSc in Computer Science from ETH Zurich (2022) with a focus on Machine Perception and Deep Learning — and further conducted research during his Master’s Thesis on Performance-Based Generative Design. Before joining ETH Zurich, he graduated with an BSc in Informatics from the University of Zurich (2019).
In the past, he worked as a Full-Stack Software Engineer (2017-2018) at BSI and was further involved (2020-2022) as a Research Assistant at the Chair for International Relations and Political Economy (IRPE) at the Department of Political Science at UZH, focusing mainly on Natural Language Processing (NLP) problems using Large Language Models (LLMs) and Transfer Learning. Further, he was a Research Assistant in the field of Automated Surgery Planning within the Research in Orthopedic Computer Science (ROCS) group at Balgrist University Hospital (2022) during his civilian service in Switzerland. He also worked on various smaller side projects on his own and has been on a journey with Tinystudio, a company he co-founded back in 2012.
Besides all this, Martin enjoys being outdoors in nature, doing sports such as hiking or climbing, watching arthouse cinema, and mixing music he finds interesting as a DJ. He is further trying to step into photography as well and you can find a sparse selection from his photographic archive on Instagram. He can also be found on Twitter, LinkedIn, and Telegram. If you want to reach him in a timely manner, good old email usually works best.
Education
-
PhD Student— Stanford University
09/2023 – Present
Research at the interesection of Deep Generative Modeling, Computer Vision, Representation Learning, Building Information Modeling, and Computational Design.
-
MSc — ETH Zurich
Computer Science
2019 – 2022
Focus on Computer Vision, Deep Learning, Machine Perception, Machine Learning, and Deep Generative Modeling.
-
BSc — University of Zurich (UZH)
Informatics
2015 – 2019
Broad curriculum ranging from Software Engineering and Economics over to Physical Geography and Foundational Courses in Computer Science.
Work
-
Teaching Assistant — Stanford University
Gradient Spaces Lab
04/2024 – 06/2024
Course: CEE 342 — Designing for Gradient Spaces. More information about the course can be found on our course website here
-
Research Assistant — ETH Zurich
Circular Engineering for Architecture (CEA)
01/2023 – 09/2023
Working with the CEA Lab on Large-Scale Spatio-Temporal Reconstruction and Volume Estimation problems using LiDAR data and Photogrammetry methods. Supervised by Prof. Catherine De Wolf
-
Research Assistant — Balgrist University Hospital
Research in Orthopedic Computer Science (ROCS)
10/2022 – 12/2022
Swiss civilian service in the ROCS group. ROCS is involved in various research areas related to the planning and execution of surgical procedures using Artificial Intelligence (AI) and Augmented Reality (AR). Supervised by Prof. Philipp Fürnstahl
-
Teaching Assistant — ETH Zurich
Chair of Innovative and Industrial Construction
02/2022 – 07/2022
Course: Introduction to Visual Machine Perception for Architecture, Construction, and Facility Management (ACFM), given at the Department of Civil, Environmental and Geomatic Engineering (D-BAUG).
-
Research Assistant — UZH
Chair for International Relations and Political Economy
02/2021 – 01/2023
Applied Natural Language Processing (NLP) problems using Large Language Models (LLMs), Semantic Analysis, Information Retrieval, Transfer Learning, and Representation Learning. My research was part of the DISINTEGRATION project, funded by an ERC Consolidator Grant. Supervised by Prof. Stefanie Walter
-
Software Engineer
BSI Business Systems Integration AG, Zurich, CH
09/2017 – 01/2018
Four month internship as a Software Engineer working mainly with Java EE, PostgreSQL, Spring Boot, and the Eclipse Scout Framework. Further continued to work part-time beside my BSc studies.
-
Co-Founder & Principal Partner
Tinystudio Bucher & Emmenegger
2012 — present
Co-founded Tinystudio, a digital agency for web projects, back in 2012. Still doing some projects with clients in Switzerland from time to time up until today.
Projects
-
rhymegen.xyz — Rhyme Generator
March 2023
Small sunday afternoon side project. A web-based client for generating badass rhyming couplets in any language: rhymegen.xyz. Powered by a strong LLM with careful prompt engineering.
-
MSc Thesis at ETH Zurich
April 2022
Title: “Performance-Based Generative Design for Parametric Modeling of Engineering Structures Using Deep Conditional Generative Models”. In collaboration with Design++. Advisors: Dr. Michael Anton Kraus, Dr. Romana Rust, Prof. Siyu Tang.
-
Menzha — watchOS App
February 2020
An App for watchOS to get the menus from the several cafeterias (i.e. mensas) around campus at ETH Zürich, UZH, and ZHdK right on your wrist. It's called Menzha.
-
BSc Thesis at UZH
May 2019
Title: “GENEVIZ: A Visual Tool for the Construction and Blockchain-based Validation of SFC Packages”. Paper (GECON 2019). GitHub Repo.
-
Imhotepic — Multiplayer Board Game
June 2016
Imhotep, the Kosmos board game of the year 2016, was implemented as a multiplayer browser game together with five people at UZH. Here is a short Demo from a former recording.
-
Hack Zurich
2016 / 2017 / 2018
Participated three times in a row at Hack Zurich, one of Europe's largest hackathons, together with Michael Ziörjen and Pascal Emmenegger. We got into the top 25 teams in 2017 and had the chance to present our iOS prototype, PeakPass, in front of the whole crowd.