Damon Lau

Principal Research Scientist, Autodesk Research

AI for how buildings are designed & built

Columbia GSAPP · UC Berkeley CED

About

Damon Lau is a Principal Research Scientist at Autodesk Research, where he leads research on AI and generative methods for designing and building the physical world, from sustainable Architecture to aircraft components.

His work starts inside real industry problems: how designers and builders actually work, where their workflows fall short, and which constraints a solution must respect. From there he builds new computational approaches (agentic systems, multi-modal AI, and generative optimization with physical simulation) and carries them into real-world use. Examples include a generatively designed, metal 3D-printed Airbus A320 partition that is 45% lighter than the conventional part yet passed the dynamic structural tests required for flight; a multi-modal sensing approach that assesses existing buildings for material value and supports recovery through robotic deconstruction; and a multi-modal agentic system, built with WZMH Architects, that automates analysis of field conditions against construction drawings and design intent.

Trained in architecture at Columbia University GSAPP and UC Berkeley College of Environmental Design, he approaches AI research with a designer’s habits: question the brief, synthesize across disciplines, and resolve competing requirements into innovative solutions.

Outside of work, he enjoys trad climbing in upstate New York, ice climbing frozen waterfalls, and backcountry skiing.

• ORCID
• LinkedIn
• Autodesk Research
• Email

Patents

AI & Machine Learning

• AI Agent for Automated Construction Issue Detection and Management — In Prep. Multi-modal large language model agent that autonomously identifies, logs, and manages construction site issues.
• Model Context Protocol for Multi-Modal Construction Data and LLM Reasoning in AEC — Filed · 2025. Standardised protocol enabling large language models to reason over heterogeneous construction data sources in AEC environments.
• Techniques for Analyzing Vehicle Design Deviations Using Deep Learning with Neural Networks — Issued · US 11,468,292 · US 11,842,262 · EP 202019006019. Deep learning approach for detecting and measuring styling and engineering deviations between target and actual vehicle design realisations.
• AI with Limited and Multimodal Data to Predict Non-Visible Materials in Existing Walls — Filed · 2025. Multimodal AI inference of hidden wall materials through a graph-based representation, enabling non-destructive building assessment.
• Techniques for Using Computer Vision to Track Production Processes for Decision Making and AI Training — Filed · 2025. Comprehensive computer vision solution for tracking construction and fabrication sequences using knowledge graphs, generating labelled data for downstream AI models.
• Techniques for Using Knowledge Graphs to Automatically Complete Draft Structural Designs — Filed · 2024–2025 · PCT + US. AI-driven BIM completion that automatically infers and populates missing structural elements from partial design data using knowledge graph representations.
• Automated Generation of Responses via Retrieval-Augmented Generation Based on Multimodal Construction Documents — Filed · 2025 · 3 continuations. RAG pipeline that retrieves and synthesises answers across heterogeneous construction document corpora including drawings, specifications, and reports.
• Techniques for Integrating Heterogeneous Construction Data into Knowledge Graphs — Filed · 2026. Methods for fusing diverse construction datasets — drawings, specifications, sensor data — into unified, queryable knowledge graph structures.
• Semi-Autonomous Self-Enriching Semantic Operational Knowledge Base — In Prep. AI-automated knowledge base with zero-configuration embedding management, multi-entity semantic linking, and semantic gap detection.

Generative Design for Buildings

• Creating Viable Building Designs on Complex Sites — Issued · US 12,073,351. Computational methods for generating architecturally viable designs on geometrically complex or constrained urban sites.
• Optimizing Building Design and Architecture for Sustainability Certification — Issued · US 11,875,296. Automated computational optimisation of building designs targeting sustainability rating thresholds across multiple certification systems.
• Optimizing Building Design for Future Transformation and Expansion — Issued · US 11,823,110. Design optimisation methods that embed long-term flexibility, transformation potential, and expansion capacity as first-class objectives.
• Rapid Exploration of Building Design Options for Ventilation — Issued · US 11,681,971. Fast computational framework for exploring and evaluating building design variants across ventilation and indoor air quality performance.
• Techniques for Designing Manufacturing Facilities — Issued + Published · Filed 2021. Computational generative techniques for layout, flow optimisation, and accelerated design of complex manufacturing facilities.
• Generative Design for COVID-19 and Future Pathogens Using Stochastic Multi-Agent Simulation — Published · Filed 2023. Stochastic multi-agent simulation optimises spatial layouts for resilience against airborne pathogens, balancing performance with programmatic constraints.
• Techniques for Generating Architectural Site Designs Based on Carbon Considerations and Goals — Published · Filed 2025. Generative methods that balance embodied and operational carbon against landscape carbon sequestration in site and architectural design.
• Generating Building Designs While Computationally Optimizing for Work Conditions — Filed · 2020. Generative building design approach that encodes occupant work condition metrics as optimisation objectives alongside conventional spatial constraints.

Material Assessment

• Evaluating Buildings for Material Value and Recovery Using Contactless Scanning — Filed · 2025–2026. Automated multi-modal contactless scanning approach to assess existing building material condition, value, and recoverability in support of circular economy deconstruction.

Earlier Work — Automotive & Industrial

• Computer-Implemented Method for Space Frame Design — Issued · US 10,606,981 · US 11,126,760 · US 11,640,487. Parametric computational method for generating optimised space frame geometries, construction kits, and structural assemblies.
• Generative Design Techniques for Automobile Designs — Issued · US 11,475,178 · US 12,190,022. Multi-objective generative optimisation applied to automotive form, structure, and performance trade-offs.
• Generating Building Designs That Optimize Productivity of the Building — Issued · CN ZL2020800800918 · Filed EP. Computational design method encoding occupant productivity metrics — circulation efficiency, daylight, acoustic privacy — as generative optimisation objectives.
• Techniques for Generating Comprehensive Information Models for Automobile Designs — Issued · US 11,775,709. Integrated system information model linking geometry, material, manufacturing, and performance data for automotive design.

Publications

2026

• Agentic AI for Autonomous Construction Issue Detection: A Multimodal Framework Using Model Context Protocol — Construction Research Congress 2026 · Conference Paper · 2026. An agentic AI framework, built on Model Context Protocol, that retrieves information from specifications and drawings, reads site imagery, and generates structured quality issues. In an office renovation case study it identified 9 of 10 documented issues.
• Why Construction Needs World Models — Autodesk Research Blog · 2026. Argues that construction teams need world models, AI systems that simulate spatial and temporal conditions, to anticipate constraints and test the consequences of decisions virtually before acting in the field.

2025

• A Contactless Multi-Modal Sensing Approach for Material Assessment and Recovery in Building Deconstruction — Sustainability · Journal Article · 2025. Multi-modal sensing framework for assessing and recovering materials during building deconstruction, supporting circular economy approaches in the built environment.

2023

• Designing for the Next Pandemic — Autodesk Research Blog · 2023. A generative design approach that uses multi-agent simulation to model virus transmission through air and surfaces and identify safer building layouts.
• Generative Design for COVID-19 and Future Pathogens Using Stochastic Multi-Agent Simulation — Sustainable Cities and Society · Journal Article · 2023. A generative design approach using stochastic multi-agent simulation to optimise spatial layouts for resilience against airborne pathogens.

2017

• Project Discover: An Application of Generative Design for Architectural Space Planning — SIMAUD 2017 · Conference Paper. An early application of generative design to architectural space planning, exploring constraint-driven layout synthesis at building scale.

Projects

• Agentic Digital Twin for Construction — 2026. AI agents that operate continuously on a live digital twin that perceives and reasons about construction progress and issues, supporting more data-informed project decisions.
• Construction AI · World Models — 2026. AI that simulates how a project will evolve, so teams can test decisions in virtual environment before committing in the field.
• Visual Reasoning for Construction — 2025. Vision-language and segmentation models convert site observations into a knowledge graph that an AI agent can reason over, turning raw photos and video into structured insight on progress, quality and risk.
• WZMH Architects — 2025. Multi-Agent construction field analysis and quality assessment, enabling automated analysis of field conditions against construction drawings and design intent.
• AI and Robotic Fabrication for Material Circularity — 2025. Paper: A Contactless Multi-Modal Sensing Approach for Material Assessment and Recovery.
• Large Scale CFD — 2023. Natural ventilation for modular housing development using novel fluid dynamics model.
• Building Auto-Complete — 2024. AI powered BIM learning from using model custom trained on historical project data, generating construction ready BIM models.
• Designing for the Next Pandemic — 2023. A generative design workflow that tests office layouts for virus transmission risk, pairing thousands of generated floor plans with a multi-agent simulation of occupant behavior and virus spread. The best layouts cut airborne risk by about 20% against the reference designs.
• Project Phoenix, Factory_OS — 2023. 316 factory-built affordable homes in West Oakland by MBH Architects, Factory_OS and Autodesk, targeting about half the typical cost, time and carbon.
• Storefront for Art and Architecture — 2019. Blending advanced fabrication with natural building materials for a gallery exhibition in NYC.
• Airbus Generative Factory Layouts — 2019. AI optimized factories for Airbus engines and wings in UK and Germany. Generative system evaluated and optimized for worker productivity, minimizing travel distances, congestion, and JIT layouts.
• Airbus VTP — 2017. Generative design applied to the leading edge structure of the A320’s vertical tail plane, evaluating tens of thousands of design iterations against stiffness, stability and mass targets, with bird impact simulated using explicit non-linear FEA.
• Project Discover — 2016. A generative design workflow for office space planning that searches parametric layouts with a multi-objective genetic algorithm across six performance goals.
• Autodesk Toronto MaRS Generative Design Office — 2016. Autodesk’s Toronto office in the MaRS Discovery District, the first large-scale generatively designed office space.
• Airbus Bionic Partition — 2015. A generatively designed, metal 3D printed structural component for the A320 cabin, 45% lighter than conventional part, passing dynamic structural tests necessary for flight.
• Mars 3D-Printed Habitat Challenge — 2015–2019. NASA’s Centennial Challenge to 3D-print habitats for Mars and beyond from local and recycled materials.
• MoMA PS1 Hy-Fi — 2014. A 40-foot tower of 10,000 compostable mycelium bricks, winner of MoMA PS1’s 2014 Young Architects Program.
• Columbia University: Advanced Data Visualization Project — 2014. Cross dicipline data visualization initative between Architecture, Computer Graphics, Biology, Astronomy departments at Columbia University.
• Visualization Marathon: Ecubelibrium — 2013. Data visualization project inventing 3d graphic mechanism illustrating the delicate equilibrium of developing economies.
• Center for Urban Farming Competition — 2013. Design for urban sustainable farming in Brooklyn, NY. 3D printed forms, customized to specific plants provide locally sourced produce to community.
• City of the Future San Francisco 2100 — 2008. Architecture competition imagining San Francisco 100 years in the future. We return the built environment to nature while residents of SF build homes in sustainable climate resiliant megastructures. IF Architecture featured in San Francisco Chronicle, January 21, 2008.