Founding Data Engineer
Confirmed on the employer's own hiring board on Sep 27, 2026. First seen by Alion on Sep 22, 2026.
TL;DR Davis is hiring a Founding Data Engineer to build the data behind our foundation model, trained from scratch to generate buildings as geometric graphs. You will own the data end to end, from raw floorplans and synthetic generation to a canonical representation, curation, large-scale pretraining and the ablations that tell us which data actually moves the model. Here the dataset is part of the algorithm.
About Davis
Davis is an AI-native real estate company accelerating early-stage development and architectural design. Today developers coordinate four to five fragmented stakeholders over weeks or months. Soon they will need only one: Davis.
We turn every input that shapes a development decision into decision-ready outputs: investor-grade feasibility studies, investment analysis, and architect-certified designs, delivered in days. Every stage pairs our proprietary AI systems with expert review, so velocity never comes at the cost of reliability.
We closed a $5.5M pre-seed co-led by Heartcore Capital and Balderton Capital, with Yellow, Evantic and Entrepreneur First, alongside angels from the founding teams of Spacemaker, Black Forest Labs, Hugging Face, Supabase, Cleo and Spore Bio. We already work with leading developers and expect to support hundreds of projects over the coming year, deepening our research, our hiring, and our coverage of the development process end to end.
The Role
You will own the data our foundation model learns from, a model we train from scratch to generate buildings as geometric graphs. Part of the corpus comes from real floorplans as images and PDFs that have to become clean, standardized graphs. A large part will be synthetic, procedurally generated building graphs, geometry and rendered floorplans, with controlled variation in style, scan noise, annotations and furniture, each kept with its ground-truth graph automatically.
You will think about the whole loop, from raw and synthetic data to a canonical structured representation, curation and validation, the training dataset, large-scale pretraining, evaluation, data ablations, and back to improving the generator and the data mixture. You are senior enough to architect the data stack and set the data strategy, and hands-on enough to write the pipelines, run the experiments, train the models and do the ablations yourself.
What you will be working on
Corpus from raw sources. Turn real floorplans (images, PDFs, scans) into clean, standardized building graphs, with the geometry and semantics that make them trainable.
Synthetic data generation. Explore strategies to expand the dataset with synthetic data.
Canonical representation and curation. Define the standardized representation, then filter, deduplicate, quality-score and validate at scale, with versioning, provenance and lineage.
Pretraining data and mixtures. Assemble the training datasets, design the mixture and the curriculum, and blend synthetic and real data for large-scale pretraining.
Data ablations. Train models to learn which data actually helps, read the results, and feed them back into the generator and the mixture.
Evaluation. Build the eval harness (datasets, metrics, regression tests, monitoring) that tracks data and model quality over time.
What We Are Looking For
You have built the dataset, not just trained on it. You have personally built or generated the data used for a large pretraining run, from raw or synthetic sources, rather than only training on a dataset someone handed you.
Senior and deeply hands-on. At least 5 years of strong experience, senior enough to architect the data stack and set strategy, but still coding the pipelines, running the experiments and doing the ablations yourself.
Data as a first-class problem. A track record where the data itself is the object: curation, filtering, deduplication, quality scoring, mixtures, synthetic generation.
Strong engineering. Deep Python, clean and typed code, async and concurrency, distributed data pipelines, TDD culture.
Nice to have
Computer vision and document understanding, images to structured output, OCR, layout extraction, segmentation, geometry extraction, vectorization, raster to vector, image to scene graph, 3D or CAD.
Graph and structured scientific data, molecular, protein or scene graphs, meshes, CAD, BIM, 3D geometry, or relational and structured world data.
Synthetic worlds and simulation, a structured state to a simulator to a renderer to synthetic images with perfect labels, then a perception model, with an eye on the sim-to-real gap.
Foundation models from scratch, real involvement in a large pretraining run, not only fine-tuning.
Public evidence of data ownership, lead on a dataset, a Hugging Face release, a dataset card, a technical blog on your pipeline, or a talk on data curation or synthetic data.
GIS and geometry, parcels, zoning layers, projections, computational geometry or constrained optimization.
Multi-country data, heterogeneous sources, localization and varying rules.
Why Join Davis
Own the data end to end, from the generator to the pretraining mixture, as the person who defines what the model learns from.
Shape a foundation model from scratch, on a structured representation no one else is training on.
Work on genuinely hard problems, image to graph, synthetic worlds and large-scale pretraining, with your work reaching clients within days.
Competitive salary and meaningful equity, at a founding level, in an early-stage company.
Join a world-class team, a mix of AI researchers, engineers and architects backed by world-class VCs.
More information about Davis, the team and the market we're going after:
Team
Mehdi (Co-founder & CEO) grew up in a family of architects and has lived this problem firsthand. He is a repeat founder who bootstrapped his first startup at 20, and graduated from Sciences Po and HEC Paris. Amine (Co-founder & CTO) is an AI researcher from École Polytechnique who worked extensively on discrete diffusion and turned down a PhD with Google DeepMind to build Davis. They started working together in July 2025 at Entrepreneur First's first European residency, a two-month lock-in in a German castle.
Today we are a team of twelve: technical profiles from Polytechnique, ENS and INRIA alongside architects and deep real estate expertise. We are small with an extremely high bar. If you want to work deeply on hard problems and see your work reach clients within days, you are the one we need.
Why We Will Win
Real estate is a $13 trillion industry that technology has largely bypassed. The professional services that feed it (design, engineering, feasibility, permitting) represent hundreds of billions in spend that no one has seriously automated.
Proptech spent the last decade selling SaaS on the edges of these workflows. It did not work, for two reasons: no professional wants another tool to learn, and no tool can automate work that runs on expert judgment. Davis makes a different bet. We do not sell tools, we sell the work, AI-generated, expert-validated, delivered in days instead of weeks. Every project compounds our data advantage across typologies, geographies and regulatory contexts.
Why No One Has Solved Architectural Design
Real estate development bleeds time and money in architectural design loops. Architects cycle through dozens of floorplan revisions to meet regulatory and client constraints, each round taking days, each missed constraint restarting the loop. Traditional CAD and BIM tools offer zero generative capability, and parametric tools only check constraints after generation, which leaves designs that frequently break under new zoning rules or irregular sites.
Generative AI has the potential to solve this, but does not yet. Fine-tuning image diffusion models on floorplans produces layouts that look plausible but fall apart under scrutiny: hallucinated rooms, mislabeled spaces, code violations no architect would accept. Pixel-space models have no concept of what a wall is or why a corridor needs to connect two things. Compliance-guidance techniques typically require segmentation at each noisy timestep, compounding errors and making major edits impractical. These layouts may look plausible yet break building codes, or need heavy post-processing before they are usable.
This is exactly the gap Davis closes. Because our model generates on structured representations rather than pixels, it produces compliant, editable plans from the start, and the data you build feeds that model directly.
We care about who you are, not just what's on your CV.
If you are drawn to what we are building but do not meet every requirement, we still want to hear from you. Studies show that women in particular tend to apply only when they meet all of the criteria. If that is you, please do not let it hold you back. We would love to receive your application.