Applied AI Engineer

About Permidia

Permidia builds software that checks construction plans against the local building code before they reach the city. Builders use it to catch problems before they submit, and building departments use it to give their examiners a clear, cited first pass. We're a small team based in Houston, Texas.

• Own the quality of the information we extract from construction drawings: doors, stairs, rooms, schedules, and the notes that tie them together.
• Build and maintain the evaluation suite that measures extraction quality on real plan sets, and keep it honest.
• Design and run experiments with vision-language models and OCR, and decide which changes ship based on the results.
• Improve how we combine what different parts of the pipeline see into a single, reliable picture of a plan.
• Help build our labeled datasets, including labeling drawings yourself when needed.
• Work with experienced plan examiners to understand what matters in a review and what a false alarm costs a builder.

• You've built extraction, evaluation, or machine learning systems on messy real-world documents, and you know the data matters more than the model.
• You think in terms of precision and recall, and you can explain the tradeoff to someone who has never heard the terms.
• You treat model changes as experiments with a clear success bar, not as upgrades.
• Strong in Python or Type Script. Experience with OCR pipelines or vision-language models is a real advantage.
• Patient with detail. Labeling data and reading drawings are part of the job, not beneath it.
• Based in the United States and able to travel to Houston occasionally.

• Health insurance: medical, dental, and vision.
• Remote-friendly, with an office in Houston.
• Paid time off.

• $150,000 to $200,000 base salary, depending on experience.
• Remote anywhere in the United States, or based in Houston, Texas.
• Full-time.