Senior Computer Vision & Machine Learning Engineer

Buzz is looking for a Machine Learning Engineer to advance computer vision initiatives for power grid infrastructure analytics. The role involves bridging research and production, adapting novel algorithms into deployed models for tasks like equipment defect detection, thermal anomaly identification, and vegetation encroachment monitoring. You will own end-to-end projects from problem framing to deployment, work within a team of experienced ML engineers, and operate with a high degree of autonomy.

Responsibilities

• Own and deliver end-to-end computer vision projects focused on:
• Equipment defect detection
• Thermal anomaly identification
• Vegetation encroachment monitoring
• Surveillance of closed areas for human and animal intrusion
• Scope, plan, and execute your own projects from problem framing through production deployment and monitoring.
• Deliver on client projects, translating client requirements and raw data into working computer vision solutions.
• Contribute to shared team projects, coordinating with other engineers to deliver against common milestones.
• Stay current with ML/CV research, identify promising methods, and evaluate their applicability to our domain.
• Adapt and implement algorithms from papers, validating against baselines and benchmarking for production viability.
• Bring the latest advances in deep learning and generative AI to bear on model training, accuracy, and reliability.
• Design and execute experiments with systematic hyperparameter tuning, ablation studies, and appropriate baselines.
• Perform structured error analysis: categorize failure modes (false positives, missed detections, localization errors, misclassifications) and break down performance by data slices (object size, occlusion, image quality).
• Select and justify model architectures based on task requirements, latency, and accuracy tradeoffs.
• Develop production-grade Python libraries for the complete ML lifecycle.
• Design and implement data pipelines including ingestion, preprocessing, annotation workflows, and quality monitoring.
• Own experiment tracking and model versioning: configurations, random seeds, dataset versions, environment specs, and model checkpoints.
• Build model serving pipelines that meet latency and throughput requirements.
• Conduct thorough code reviews and write integration tests for ML pipelines.
• Share knowledge with teammates and contribute to best practices for model development, evaluation, deployment, and monitoring.
• Advocate for and uphold software quality standards within the ML team.
• Communicate research findings, technical decisions, and model limitations clearly to stakeholders and clients.

Requirements

• 5–10 years of industry experience in computer vision and machine learning.
• Deep expertise in modern computer vision and deep neural networks, including:
• Object detection
• Semantic segmentation
• Image classification
• Vision transformers and foundation models
• Vision language models
• Similarity search
• Proven track record of deploying and maintaining ML models in production.
• Experience selecting, fine-tuning, and adapting model architectures (CNNs, transformers, foundation models) for specific use cases.
• Demonstrated ability to read ML research papers, extract the key ideas, and implement them.
• Ability to debug training instabilities and conduct systematic error analysis.
• Proficiency in Python and the core ML stack:
• PyTorch and Lightning
• OpenCV
• NumPy and pandas
• Scikit-Learn
• FastAPI and Pydantic
• Strong software engineering practices, including:
• Git version control
• Unit and integration testing (Pytest)
• CI/CD pipelines (GitHub Actions)
• Docker and reproducible environments
• Experiment tracking and model versioning
• ML DevOps
• Python type hinting
• Proven ability to own technical projects independently, from problem framing through production deployment.

Nice to have

• Multi-modal computer vision
• Custom object detection model development
• Generative models for data augmentation
• ML deployment on edge devices
• Extracting measurements from GIS and/or drone metadata enriched imagery
• Model quantization
• Systematic hyperparameter tuning