仕事内容
職務内容
H2 Corporation is an AI-native company modernizing one of the largest and most underserved industries in the world: construction.
Founded in Tokyo in 2020 and now expanding globally, H2 is building the next generation AI-powered system of record for quoting and orders for the $13T global construction industry. Our AIS Platform automates takeoff, estimation and quoting, and orders workflows for the largest construction enterprises, helping the industry address a generational labor shortage with AI-driven productivity. We are backed by top-tier investors including NEA, JAFCO, and Spiral Capital, and MetaProp. We are currently a team of 120 members distributed across 5 offices globally, and are scaling rapidly across Japan and the United States.
THE OPPORTUNITY:
H2 is looking for a Staff AI Engineer, Computer Vision to help define and build the AI models to automate complex takeoffs (extracting information from 2D PDF drawings) and quoting workflows.
Construction is one of the most complex visual and operational domains in the world. Drawings, field photos, estimates, measurements, schedules, and project documents are dense, messy, high-stakes, and still largely interpreted by humans. H2 is building AI systems that can understand this data, automate manual workflows, and turn construction domain knowledge into reliable software.
You will own senior/staff-caliber technical work across computer vision, multimodal AI, model evaluation, data strategy, and production ML systems. We care less about years alone and more about demonstrated ability to own complex projects, mentor others, set direction, and solve ambiguous AI problems.
This is not a research-only role, and it is not an "API wrapper" role. Most of our models are built and trained in-house and to-date, we do not wrap any of our features around third party LLMs. We need someone who can reason deeply, prototype quickly, ship production systems, analyze failure modes, and raise the technical bar for AI engineering at H2.
WHAT YOU'LL DO:
- Own the design and delivery of production computer vision and multimodal AI systems for H2's construction products.
- Lead technical approaches for construction drawing understanding, visual/document intelligence, field data, takeoff, estimating, surveying, and related workflows.
- Evaluate and apply CNNs, Transformers, diffusion models, LLMs, VLMs, segmentation models, detection models, OCR/document AI systems, and other modern architectures where appropriate.
- Define model evaluation strategy, success metrics, benchmarking methods, failure analysis, and quality thresholds for production AI capabilities.
- Build and improve data pipelines, labeling strategies, synthetic data workflows, active learning loops, and feedback systems that improve model quality over time.
- Fine-tune, adapt, and productionize open-source and foundation models while optimizing for accuracy, latency, cost, reliability, and hardware efficiency.
- Partner with product and engineering leaders to turn ambiguous customer problems into technically sound, customer-facing AI capabilities.
- Make clear technical tradeoffs across model quality, user experience, data availability, infrastructure cost, and implementation speed.
- Mentor other engineers, review technical designs, and help establish strong AI engineering practices across the team.
応募要件
MUST:
- 8+ years of software engineering, machine learning, or applied AI experience, or equivalent senior/staff-caliber ability demonstrated through complex production AI work.
- 5+ years building and deploying computer vision, multimodal AI, or production ML systems, with evidence of owning systems beyond research and experimentation.
- Strong Python skills and hands-on experience with PyTorch, TensorFlow, Keras, or similar deep learning frameworks.
- Deep understanding of computer vision, machine learning fundamentals, model evaluation, data quality, and production ML workflows.
- Experience wi