Founding AI/ML Engineer

Washington, D.C.

• Hybrid (3 days/week onsite)
• Full-time
  
  
  

Compensation: $190,000–$200,000 equity

About The Company

An early-stage (Seed) startup building the AI-powered workflow layer for the construction industry — a massive sector that has largely skipped the software era, where the core customer has seen little meaningful software innovation in over 15 years. Agents and multimodal AI are the first real wedge in, and because construction diagrams aren't represented in standard foundation-model training data, training in-house models is a core technical bet. :$2M ARR with a plan to scale 10x over the next 12 months.

Founded 2023

• :11–50 people
• Industry: Property Tech / Construction Tech
  
  
  

The Role

Own the model-development side of the platform as the founding AI/ML hire. Construction diagrams and floor plans aren't in any off-the-shelf foundation model's training data, so the company trains its own — and this role leads that work. Roughly half computer vision (the irreplaceable part) and half agentic orchestration and retrieval. Reports to the Chief AI Officer and provides technical leadership to a small existing AI/ML team.

What You'll Be Doing

• Train and improve in-house computer vision models for construction diagrams (segmentation, object detection, classifiers)
• Lead architectural and model-improvement work where off-the-shelf models hit their ceiling
• Own the model-development roadmap and set technical direction for the AI/ML team
• Build agentic orchestration and retrieval flows (RAG, fine-tuning, local vs foundation model decisions, cost optimization)
• Provide technical leadership and guidance to the existing AI/ML engineers
  
  
  

Tech stack: Python, PyTorch, CUDA; computer vision (2D segmentation, object detection, classifiers; SAM / SAM3-tier foundation models); agentic orchestration & retrieval (LLMs, RAG, fine-tuning small language models, local vs foundation model trade-offs)

Requirements

• Computer vision, 2D segmentation depth
• Python PyTorch CUDA
• Train in-house models (not just fine-tune)
• Construction diagrams / floor plans bonus
• East Coast, 3 days hybrid at Philly hub
  
  
  

Green Flags

• 2D segmentation expertise with published work in floor plan localization, raster-to-sequence, structural priors, or comparable architectural CV. Direct domain match.
• Construction diagrams, floor plans, or technical drawings personal or professional exposure. Genuinely rare and high-signal.
• PyTorch CUDA fluency with architectural model improvement track record (not just fine-tuning APIs)
• Research-trained but now hands-on.
• Agentic orchestration / retrieval experience as a secondary skill. Strong plus but not blocker.
  
  
  

Red Flags

• TensorFlow-only background without PyTorch transition. PyTorch is the expected core framework; TensorFlow is dated for this work.
• Pure backend ML engineer without CV depth.
• Off-the-shelf model orchestrator without ability to train custom models.
• Pure research without willingness to ship. The role is hands-on technical leadership, not academic.
• Cannot do East Coast hybrid (3 days/week onsite). Flex only for genuinely exceptional candidates.
  
  
  

Why Join

• Own the entire model-development side of the platform as the founding AI/ML hire, setting technical direction for the team
• Work on a genuinely hard, defensible problem — training in-house CV models where off-the-shelf foundation models fall short
• Early-stage upside: meaningful early-employee equity grant, a freshly closed seed round, and a credible 10x growth plan
  
  
  

Details

• Location: Washington, D.C.
• Work policy: Hybrid — 3 days/week onsite at the Philadelphia hub (flex for exceptional candidates)
• Compensation: $190,000–$200,000 equity
• Visa sponsorship: H-1B available
• Employment type: Full-time