• Embodied Computation Lab

    Location:Princeton, NJ
    Client:Princeton University
    Role:Initial concept design, facade design and machine learning application

    The Embodied Computation Lab is a new lab facility for the School of Architectura at Princeton University. The goal of the design was to provide a large flexible space in which the university’s researchers could experiment with new technologies such as robotics and embedded sensor systems.

    Although simple in form, the building incorporates a variety of sustainability strategies. One of the most unique is the building’s facade, which is composed entirely of reused scaffolding planks salvaged from construction sites around New York City.

    These boards are certified by OSHA to meet the structural and durability standards to be used on a construction site. However, due to the wide variety of conditions found on construction sites and the difficulty of ongoing testing they are only certified for one year, after which they are typically thrown away and replaced.

    To give these boards a new life within the Embodied Computation Lab, we worked with a local salvage company to source 2,000 of these boards for the facade of the building. Looking at the boards, we were struck by how a year spent in a variety of construction site conditions had weathered each board differently, causing the initially standardized boards to reclaim some of their natural material qualities.

    In the design of the facade, our goal was to further exaggerate and express the differences within each of the boards. To do this we partnered with local Brooklyn artist Evan Eisman, who has been experimenting with sandblasting natural woods in order to reveal their unique grain structures. Noticing that the grain structures are most unique at the knots where branches connected to the trees, we decided to focus the sandblasting only on the locations of the knots.

    Manually finding and sandblasting the knots on all 2,000 boards would have been extremely challenging and time-consuming, so to scale up the process we developed a Machine Learning application using Convolutional Neural Networks to automatically find the location of all knots in images of the boards. We could then feed this data into a custom-designed computer numerically controlled (CNC) sandblasting machine to sandblast the location of each knot.

    To train the Machine Learning model we needed a large amount of data that classified images of wood boards according to whether or not they contained knots. To generate this data we created a web application that presented a single square image of a wood board and asked the user to click a button specifying if the image contained a knot or not.

    Using this app, we crowdsourced a training set of over 10,000 classified images that we could use to train the model.

    After training, the model gave us highly accurate and robust predictions of where knots occurred in images of wood boards.

    We could then apply this model to find the locations of knots in each board, and then pass this data to our CNC sandblasting machine to treat all the boards of the facade.

    These images show the boards being installed on the building facade, and the finished facade in its context on Princeton’s wooded campus. This project shows the potential for using advanced tools of computation, Machine Learning, and robotics to work with and reveal the unique features and identity of natural materials.


    Embodied Computation Lab
  • Bionic Partition

    Role:Design lead and project management

    The Bionic Partition is a design concept for a new component for the Airbus A320 aircraft developed in collaboration with Airbus, Autodesk, and AP Works. The goal of the project was to redesign the interior partition component of the aircraft such that it performs equally well to the existing partition but weighs 50% less. Weight reduction is a critical focus of the aircraft industry today and will be the key to maintaining the environmental and economic sustainability of air travel into the future.

    To achieve this difficult goal, the Bionic Partition leverages cutting-edge metal alloys developed at Airbus, new fabrication techniques in metal 3d printing, and new design software which can iterate through many design options, analyze each one for its performance, and ‘evolve’ better performing designs over time. By combining human intuition with artificial intelligence, this novel generative design workflow can create much better and higher performing designs that would be possible through a traditional design process. After three years of development, the project resulted in a complete, testable prototype of the world’s largest metal 3D printed airplane component. This prototype is currently undergoing 16G crash testing as part of the process for certification and integration into the current fleet of A320 planes.

    The Bionic Partition was unveiled at the Autodesk University annual conference in Las Vegas in 2015. It has been covered in Wired Magazine and The Wall Street Journal, was a finalist in Fast Company’s 2016 Innovation by Design Awards and was the winner of a 2017 Architizer A+ Award for innovation in 3d printing.

    Rendering of final partition design with cover panel and cabin attendant seat (CAS)

    While the partition wall may seem like a relatively simple component, it actually presents two complex structural challenges. First, the partition must support a fold-down cabin attendant seat (CAS). Unlike the partition, the CAS is not attached to the airplane’s fuselage or the floor, thus the full weight of two flight attendants and the seat itself must be transferred through the partition into the aircraft’s structure. Second, due to new safety regulations, the partition must include a panel called the ‘stretcher flap’ which can be removed to allow a stretcher carrying a sick or injured passenger to be carried around the corner from the seating area to the galley and exit. This results in a big hole in the partition which makes it difficult to route forces from the CAS directly into the aircraft’s fuselage.

    Diagram of geometry system with 50 inputs, 2 constraints derived from FEA simulation, and two objectives derived from the geometry of the model
    Diagram of computational geometry system based on the growth of slime mold
    All designs explored during the optimization process plotted according to the two objectives. Colour represents the generation in which the design was evaluated, with blue for earlier and red for later designs. Designs with a black outline are part of the Pareto-dominant set of optimal solutions.

    Rationalization of geometry and fabrication of final prototype
    Final design component breakdown (left) and manufactured prototype (right)
    Airbus engineers test-fitting bionic partition into A320 fuselage


    • 12/01/15 – WIRED – Airbus’ Newest Design Is Based on Bones and Slime Mold
    • 01/19/16 – Architizer – Bionic Partitions: A Close-Up Look at The Living’s Pioneering 3D-Printed Structures
    • 01/21/16 – Architect Magazine – The Living and Autodesk Apply Bionic Design to an Airbus 320 Partition
    • 09/12/16 – Fast Company – Innovation by Design Awards 2016 Finalist in “Experimental” category
    • 11/29/16 – Wall Street Journal – Manufacturers Take a Page From Mother Nature
    Bionic Partition
  • 2221 Water Way

    Location: Seabrook, TX
    Client:Sandor and Natalia Nagy
    Role: Concept design, design development, and construction management

    The house at 2221 Water Way is located along the coast of the Gulf of Mexico, approximately one hour outside the center of Houston, TX. The design concept is based on a series of spatial strategies which are designed specifically to fit the clients’ lifestyle.

    The ground floor is composed of a single living space combining the and living area. In the center of this area is a double-height space spanning from the ground all the way to the peak of the roof. This tall space is accentuated by a large natural fireplace made of exposed concrete block. An open loft on the second floor creates an informal gathering area that is visually connected to the living area below. The double-height space creates a separation between two guest rooms at the front and the large master suite in the rear.

    On the other side of the house, an open garage and interior courtyard break up the house’s form, giving it an urban scale and creating a range of outdoor spaces to fit the tropical environment. These include a large covered patio in the rear, a ground-floor deck connected to a roof deck with a direct connection to the master suite, and a private balcony shared by the two guest rooms. The courtyard also creates an isolated dining room that is separate from the main living area but connected directly to the street through the garage.

    Interior view from the kitchen to the living room and second floor loft
    View from front and back yard
    View of second floor terrace into master suite
    View of master bathroom (left) and kitchen (right)
    View into interior courtyard

    Sketches of early massing, plan and elevation development
    Ground and second floor plans
    Building sections

    View from back yard

    2221 Water Way
  • Hyfi

    Location: Queens, NY
    Client:MoMA PS1
    Role:Design lead and construction management

    Commissioned by the MoMA PS1 museum as part of their annual ‘Young Architects Competition’, Hy-Fi offers a captivating physical environment and a new paradigm for sustainable architecture. In 2014, we tested and refined a new low-energy biological building material, manufactured 10,000 compostable bricks, constructed a 13-meter-tall tower, hosted public cultural events for three months, disassembled the structure, composted the bricks, and returned the resulting soil to local community gardens. This successful experiment offers many possibilities for future construction.

    To make the material for the structure, 10,000 compostable bricks were grown at a specialized facility in upstate New York. To make the bricks, a set of vacuum-formed molds were filled with a mix of mycelium roots and agricultural waste. Then, over the course of 5 days, the mycelium grows and eats the agricultural waste, eventually forming a solid brick. These bricks were then used to construct the temporary structure. At the end of the summer, the structure was disassembled and all 10,000 bricks were composted and donated to local community gardens in Queens, NY.

    A custom bottom-up agent-based algorithm was created to solve the complex problem of laying out the brick pattern to describe a doubly-curved surface from a set of only 3 unique size brick modules.

    Views of installation during summer ‘Warm-up’ parties at the MoMA PS1.
    View of interior of structure, where the removal of structurally unecessary small brick modules creates pseudo-random openings letting rays of light into the interior space.
Go Back

About me

I am a designer, developer, and entrepreneur creating technology to transform the building industries. Trained as an architect, I started my career in design offices working at a range of scales from homes in Chicago to mega-developments in China. An interest in data and computation led me to new opportunities in the tech industry, including a role as Principal Research Scientist at Autodesk. I have taught architecture and technology at several schools including Columbia University and Pratt Institute. I am now pursuing new ventures to bring emerging technologies like Generative Design and Machine Learning to the architecture, engineering, and construction industries.

Get in touch