What are the responsibilities and job description for the AI Team Lead – Computer Vision and Structural Defect Detection position at Inspekt AI?
Inspekt AI builds computer vision that replaces slow, manual façade inspections with scalable, repeatable, image-based workflows. We deploy models into real customer projects, and we’re now scaling model quality, training reliability, and AI-driven inspection throughput. You will lead the AI engineering function, responsible for our core perception stack: High-resolution façade defect detection, image quality filtering, segmentation/classification of façade components, and 2D/3D data fusion into inspection pipelines. This is not a research-only role. Your work must ship to production, be measured in live inspections, and directly improve the quality and efficiency of our customer deliverables. Team stage & growth:
- We are expanding the team quickly
- You will be the technical anchor and player-coach who sets direction, upgrades the training/data pipeline, and scales the team into a high-output AI function.
- ~80% hands-on technical work (coding, experiments, architecture, model iteration, PR reviews).
- ~20% technical leadership (mentoring, standards, roadmap input, hiring support).
- Hands-on AI / Computer Vision Delivery (~80%)
- Design, implement, and improve production-grade CV models for façade inspection, including:
- Defect detection with precision-first targets.
- Segmentation/classification of façade components and materials.
- Image quality evaluation/filtering to improve downstream inspection accuracy and reduce noise.
- Own the end-to-end model lifecycle:
- Data selection, preprocessing, and augmentation at scale (very high-res imagery, large project volumes).
- Labeling strategy and training-specific QA in collaboration with annotators and façade engineers.
- Training, validation, and evaluation using clear project-relevant metrics.
- Batch deployment to production pipelines and continuous monitoring/improvement.
- Build and maintain a systematic experimentation engine: hypotheses, baselines, ablations, and clear readouts of what worked and why.
- Write production-quality code: modular Python, robust training/inference components, tests for critical paths, and clean integration with internal services/APIs.
- Training Data Pipeline & Evaluation Foundations (Top Priority)
- Establish a high-quality AI training data pipeline that is distinct from human annotation workflows, including:
- Dataset versioning and lineage.
- Sampling strategy and coverage guarantees across projects/building types.
- Label QA rules for training fitness (consistency, edge cases, class leakage).
- Repeatable train/val/test splits and regression tracking.
- Create repeatable error-analysis workflows and dashboards tied to real project outcomes.
- Technical Leadership & Mentoring (~20%)
- Act as the go-to technical expert for AI engineers: unblock others on architecture, training stability, debugging, and performance issues.
- Set and enforce standards for AI engineering:
- Coding conventions, documentation, testing.
- Reproducibility and traceability for models and datasets.
- Experiment tracking discipline.
- Shape the AI roadmap with the technical management: recommend priorities based on impact, feasibility, and delivery constraints; clearly articulate trade-offs.
- Support hiring and onboarding of new AI engineers: interviews, technical assessment design, and structured onboarding to ramp quickly.
- Model Strategy & Architecture
- Lead strategy for the model portfolio:
- Decide when to use one generalized defect model vs. multiple specialized models (by building type, material, region, or inspection context).
- Define decision criteria and rollout plan, including how models are selected per project.
- Define and refine requirements for scalable training and inference architecture, ensuring reliability and cost-awareness.
- Collaboration & Stakeholder Management
- Work with façade engineers and delivery teams to:
- Translate domain knowledge into useful guidelines, rules, and model objectives.
- Validate that AI outputs are usable in real inspection workflows.
- Establish a tight feedback loop that drives iterations and improvements.
- Work with cloud/infra engineers to specify training/inference requirements; they build the infrastructure, you ensure it meets model needs.
- Communicate clearly with leadership on progress, risks, hiring needs, and trade-offs.
- 5 years hands-on ML/DL experience, with 3 years in computer vision.
- Strong experience in image detection and segmentation using modern architectures (e.g., YOLO family, Mask R-CNN, UNet, transformer/ViT-based models).
- Proven track record taking CV models from prototype to production used in real projects and iterating based on monitoring error analysis.
- Strong software engineering fundamentals:
- Python, PyTorch/TensorFlow/JAX.
- Clean, maintainable codebases; testing for critical paths; CI/CD literacy.
- Demonstrated ability to lead technically:
- You’ve been a senior reference point, reviewed others’ work, guided technical direction, and mentored engineers.
- Comfortable defining success metrics from ambiguity and defending trade-offs with data.
- Bias toward shipping and measurable customer impact over shiny research.
- Experience with very high-resolution imagery (40–60 MP) and large image volumes.
- Drone imagery, mapping, photogrammetry, or geospatial workflows.
- Building/infrastructure inspection, civil engineering, or similar domains.
- Familiarity with MLOps tooling (MLflow, W&B, SageMaker, Vertex, or equivalent).
- Experience building tooling/workflows for annotators, QA teams, or domain experts.
- A fully remote position, allowing you to work from anywhere in the Philippines
- Competitive salary and benefits package (PTO and HMO)
- Employee Stock Ownership Plan (ESOP) eligibility
- Flexible working hours to accommodate project needs and time differences.
- Opportunities for professional growth and development in a company at the forefront of AI-driven building inspection technology.