Machine Learning Engineer

AI Engineer — Multimodal (Vision + NLP), Civil Infrastructure Inspection

Role Description

Dynamic Infrastructure builds AI that helps civil engineers use inspection images/video and technical reports to assess asset condition, predict deterioration, and prioritize maintenance for bridges, roads, and other critical infrastructure. This is a full-time, hybrid role based in Emek Hefer (Yarok Center), Israel.

You’ll move quickly from idea to working system—designing and implementing complex vision-language architectures that turn messy, real-world inspection data into structured defect findings and maintenance insights. The work requires a strong understanding of modern AI literature, including how these models behave at inference time, their tradeoffs, and common failure modes. The focus is implementation: prototyping, running experiments, and hardening what works into reliable production pipelines, in close collaboration with civil engineering experts.

What You’ll Work On

• Complex Vision Pipelines: Multi-stage CV pipelines (classic + deep learning) for defect signals from large-scale images/video—detection, classification, matching/registration, etc.
• Domain-Specific NLP: Extract key information from technical engineering reports across messy formats and niche terminology, including using foundation models to parse and normalize complex documents.
• Hybrid Architectures: Use deep models and classic CV/ML alongside foundation models by converting raw inputs (images, text) into reliable intermediate signals that help the overall system handle difficult cases and produce accurate, engineering-grade outputs.
• RAG + Agentic Workflows: Build retrieval/agent flows that query inspection media, reports, and metadata to generate structured client outputs.
• Productionization & Scaling: Turn prototypes into cloud-based, GPU-heavy pipelines that are stable and maintainable.
• Domain Expert Collaboration: Work closely with civil engineers on labeling, evaluation, and real-world failure analysis.

Role Breakdown

• 40% AI Systems / Pipeline Engineering: Building modular pipelines and agentic workflows.
• 30% Research and Prototyping: Exploring model architectures, multimodal learning, and experimental evaluation.
• 20% Data Handling: Preparing training data, managing multimodal datasets, evaluations, and failure analysis for model improvement.
• 10% Cloud & Product Integration: Deployment of AI systems and production support.

Qualifications

Must-haves

• Hands-on experience shipping ML features from prototype to production—building data pipelines, integrating model inference (batch or real-time), and running evaluation through deployment and iteration.
• Python proficiency for ML research and production-grade development (data analysis, inference, evaluation, and tooling).
• Hands-on experience in either Computer Vision or NLP (applied ML, not just coursework).
• Solid software engineering practices: maintainable code, reproducibility, testing/validation mindset, and debugging ability.
• Understanding of modern deep learning architectures and practical limitations/failure modes on real-world data.

Experience & Education

• 2+ years of hands-on applied AI/ML engineering experience.
• MSc in Computer Science/Engineering (preferred) or equivalent practical experience.

Strong Advantages

• Experience spanning both CV and NLP, or building multimodal vision-language systems.
• Experience training/fine-tuning models (running training jobs, hyperparameter tuning, experiment tracking).
• Experience with RAG and/or tool-using (agentic) LLM workflows.
• Familiarity with Docker, AWS, Terraform, and CI/CD pipelines.
• Experience running deep learning workloads on cloud GPUs.