Post-Training Lead, Vision-Language AI for Construction

ironsite-ai; San Francisco (CA); On-site; USD 250,000 - 400,000; Full time

Benefits offered by this job

Health, dental, vision

401k + 6% match

Significant equity

GPU compute access

Job summary

Ironsite is seeking a Post-Training Lead to own the end-to-end post-training effort for state-of-the-art Vision-Language Models tailored to construction sites. You will train models, guide SFT and RL studies, and mentor researchers while scaling experiments for production.

You will work in a fast-paced startup environment, shaping baselines, long-context architectures, and benchmarks to forecast real-world jobsite outcomes with practical impact.

Qualifications

• Background in Computer Science, ML, AI, or Robotics.
• Strong Python programming skills and software engineering basics.
• Experience with large-scale vision and/or language datasets.
• Hands-on experience with video data and temporal reasoning.

Responsibilities

• Architect & Train Novel VLMs: Design, train, and iterate on general-purpose Vision-Language Models fine-tuned for spatial intelligence in the construction site.
• Own the Post-Training Roadmap: Lead research goals, baselines, post-training recipes (SFT/RL), long-context architectures, and visual reasoning.
• Build on the Construction Intelligence Benchmark: Expand video QA, temporal reasoning, activity recognition, and site-level analytics.
• Build Scalable Pipelines: Create and own model training/evaluation pipelines for rapid experimentation and deployment.
• Provide Technical Leadership: Set direction for post-training, review designs/PRs, mentor researchers, uphold experimental rigor.

Skills

Python

PyTorch

JAX

Robotics

Machine Learning

Vision-Language Models

Video data

MLOps

Tools

PyTorch

JAX

MLOps tools

Job description

Ironsite is seeking a Post-Training Lead to own the end-to-end post-training effort for state-of-the-art Vision-Language Models tailored to construction sites. You will train models, guide SFT and RL studies, and mentor researchers while scaling experiments for production.

You will work in a fast-paced startup environment, shaping baselines, long-context architectures, and benchmarks to forecast real-world jobsite outcomes with practical impact.