AI Research Scientist

MeltPlan | Planning Engine for the Built Environment

MeltPlan is building the “planning engine” for the $14 Tn construction industry, an AI system designed specifically to optimize decisions before construction begins. While design software optimizes use and aesthetics and construction software optimizes execution and control, MeltPlan is building the missing layer - software that optimizes decisions and tradeoffs upstream, before scope is locked, procurement begins, and change orders become inevitable. MeltPlan’s long-term goal is to help teams make construction “boring” by making planning more intense: surfacing constraints and tradeoffs early, aligning stakeholders before plans are frozen, and reducing the need for late-stage redlines, rework, and change orders.

MeltPlan is founded by operators who have built at scale. Kanav previously co-founded Innovaccer, a $3Bn healthtech company focused on making US healthcare more affordable and accessible. He’s now applying that systems-level thinking to construction.He’s joined by Tanmaya Kala, former Project Executive at DPR Construction, who led large commercial, healthcare, and life sciences projects. We combine deep tech scale with real construction execution.

What This Role Really Is

We are looking for an AI Research Scientist – Computer Vision to enhance and manage the PlanGraph model, which transforms 2D drawings into structured graphical representations of building elements. The role involves solving downstream business use cases such as quantity takeoff, code compliance, value engineering, and constructability analysis.We are specifically looking for hands-on researchers with experience in solving real-world Computer Vision problems and building custom vision models, VLMs, or VLLMs for production-grade applications.

What You’ll Do

• Build and optimize custom Computer Vision models, VLMs, and VLLMs for construction intelligence workflows.
• Solve downstream business use cases including quantity takeoff, code compliance, value engineering, and constructability analysis.
• Conduct applied research in Computer Vision, multimodal AI, document understanding, and spatial reasoning.
• Design, train, fine-tune, and evaluate production-grade AI models using large-scale datasets.
• Improve model accuracy, robustness, inference performance, and scalability for real-world deployment.
• Collaborate closely with engineering and product teams to convert research into deployable AI systems.
• Develop benchmarking, evaluation, and validation frameworks for model performance measurement.
• Stay updated with state-of-the-art research in Computer Vision, Vision Transformers, multimodal AI, and Generative AI.
• Contribute to research documentation, patents, publications, and internal technical knowledge sharing.
• Work on real-world construction datasets, drawings, and complex edge-case problem solving.