Resume — Clay H. Seifert

Product engineer and computational designer building ML systems at the intersection of architectural geometry and real-world design workflow. Proven track record deploying production systems — Archipedia.ai (multi-modal AEC precedent retrieval) and a WebGPU finite element solver — grounded in deep AEC domain knowledge and published research in geometric representation. Experienced translating architect workflows into product requirements and technical systems. Focused on making complex computational tools accessible, useful, and durable in practice.

Experience

Drafted AI

San Francisco, CA · 2026–Present

Product Engineer — Founding Team (Employee #6)

Building core editing and interaction features for an ML-driven architectural design platform (raised $17M at $100M valuation).
Leading UI/UX design for generative workflows; translating architect needs into product specs and engineering requirements for a technical team.
Bridging AEC domain logic, ML systems, and user-centered product development at the earliest stage of the company.
Contributing to product strategy including roadmap decisions at the intersection of computational design and ML capability.

Testfit

Dallas, TX · 2023

Product Development Analyst

Reported to CEO/VP Product; translated architect workflows into product requirements and technical specifications for a 20+ person engineering team.
Supported product R&D across site planning and feasibility workflows; contributed to shipped features and internal documentation.
Built communication infrastructure between AEC domain logic and software engineering execution.

Rick Wright Architects

Dallas, TX · 2020–2023

Architectural Intern

Design and documentation across single-family residential, commercial, industrial, and institutional project types.
Gained firsthand experience with conventional AEC practice tools and their constraints — the foundation for my computational design research agenda.

Selected Projects

Archipedia.ai

2025–Present

Multi-Modal AEC Precedent Retrieval System

Built and deployed a production search engine for architectural precedents supporting query by geometric constraints, material assemblies, and spatial logic — active deployment at archipedia.ai with a growing practitioner user base.
Developed end-to-end ML pipeline: vision embedding models (CLIP/ViT), vector similarity search (FAISS), multi-modal indexing, and an interactive React/FastAPI interface.
Designed the retrieval system to treat archives as traceable design collaborators rather than aesthetic inspiration.

PyTorch · CLIP/ViT · FAISS · React · FastAPI · Python

GPU-Accelerated Structural Analysis

January 2026

Real-Time FEA Solver — collaboration with Lucas Epp, StructureCraft

Developed a real-time finite element analysis solver for 2D plate structures using WebGPU compute shaders — motivated by the need to integrate structural reasoning directly into generative architectural design workflows.
Achieved sub-20ms solve times for 100,000 degrees of freedom with arbitrary polygonal boundaries, internal holes, and mixed support conditions.
Implemented multigrid iterative methods with real-time stress visualization, moment/deflection computation, and isocurve rendering.
Demonstrated that browser-based scientific computing can function as a viable interaction paradigm for constraint-aware generative design.

Research

Latent Geometric Representations in Architectural Fragments

2025–Present

Independent Research

Developed Vision Transformer–based methods for recovering latent generative operations (extrusion, revolution, truncation, boolean operations) from architectural imagery — making fabrication logic computable from visual evidence.
Created patch-aware retrieval workflows enabling evidence at the level of geometric parts and operations rather than whole-image similarity.
Investigated geometry as proxy for fabrication constraints and assembly logic — directly relevant to computational design tooling in a production studio context.
Published at AAG 2025, ACADIA 2025, CAADRIA 2026; submitted to ACM DIS 2026.

Vision Transformers (ViT) · UMAP / t-SNE · similarity search · clustering

Learning Agents Research Group, Texas Robotics

Austin, TX · 2024

Embodied Interaction Designer (Advisor: Justin Hart)

Designed robot morphology and interaction affordances for human-robot interaction in real-world environments; translated research objectives into CAD geometry and fabrication-ready constraints.

Publications

Seifert, C., Guerra, D., Kokan, A., and Danahy, P. (2026). Multi-Modal Precedent Retrieval: Patch-Aware and Tri-Scalar Rank Weighted Guidance for Design Search. CAADRIA 2026.

Seifert, C. (2025). Patterns Across Languages: Analyzing & Mapping Latent Projective Geometry in Cross-Cultural Architectural Fragments. ACADIA 2025.

Seifert, C. and Danahy, P. (2025). Synthetic Vision: Analyzing Latent Projective Geometry in Photogrammetric Fragments. AAG 2025.

Entrepreneurship

Resia AI

Austin, TX · 2023–2025

Founder

Co-founded home renovation ML platform; led product validation and team assembly (10+ members); secured $20K in grants; participated in SEAL 2024 accelerator (Techstars Network).
Pitched to 50+ investors; learned rapid iteration, stakeholder interviewing, and product scoping under resource constraints.

Education

The University of Texas at Austin

Austin, TX

Bachelor of Architecture — Dual Focus: Architecture & Computer Science

Technical Skills

Comp. Design: Grasshopper, Rhino.Python, parametric modeling, shape grammars, generative systems, geometry processing, mesh operations, assembly representations
ML / AI: PyTorch, TensorFlow, Vision Transformers (ViT), CLIP, embedding models, similarity search (FAISS), dimensionality reduction (UMAP, t-SNE)
Languages: Python, JavaScript / TypeScript, C++, Bash / Shell
Web / Systems: React, Node.js, FastAPI, WebGPU, Three.js, browser-based scientific computing
Sci. Computing: NumPy, SciPy, finite element analysis, numerical optimization, parallel computing, GPU acceleration
Design Tools: Rhino, Grasshopper, AutoCAD, Revit, Adobe Creative Suite, Blender
Fabrication: CNC milling, laser cutting, 3D printing, physical prototyping, assembly constraints

Other

Marathon runner · Muay Thai (Woodenman) · Brazilian Jiu-Jitsu · Freetail Hackers