Key Responsibilities
Computer Vision Pipeline Development
- Design and implement real-time CV pipelines for object detection, tracking, and classification meeting <100ms p99 latency SLOs
- Build multi-object tracking systems across camera feeds with re-identification and trajectory forecasting
- Develop preprocessing pipelines for video streams (frame extraction, normalization, augmentation) with error handling and backpressure mechanisms
- Implement annotation workflows and active learning loops to continuously improve model quality
Model Engineering & Optimization
- Fine-tune and evaluate SOTA open-source models (YOLO, EfficientDet, DETR families) on domain-specific datasets
- Optimize inference throughput: batching strategies, model quantization (INT8/FP16), ONNX/TensorRT conversion, and multi-GPU orchestration
- Build A/B testing frameworks to measure model performance (mAP, FPS, recall@IOU) in production
- Maintain model registry with versioning, lineage tracking, and rollback capabilities
Production ML Infrastructure
- Architect scalable ML services exposing REST/gRPC APIs with authentication, rate limiting, and circuit breakers
- Containerize models and services (Docker) with CI/CD pipelines for automated testing and deployment
- Implement monitoring dashboards tracking inference latency, GPU utilization, prediction confidence distributions, and data drift
- Own incident response: debug production issues, conduct root-cause analysis, implement permanent fixes
Software Engineering Excellence
- Write maintainable Python code with type hints, unit/integration tests (pytest), and API documentation
- Design clear data contracts between services; validate schemas with Pydantic/protobuf
- Conduct thorough code reviews focusing on performance, maintainability, and ML best practices
- Document system architecture, model cards, and operational runbooks
Collaboration & Mentorship
- Partner with data engineers on annotation tooling, dataset pipelines, and feature stores
- Work with DevOps to optimize Kubernetes deployments, autoscaling policies, and cost efficiency
- Mentor junior engineers on CV fundamentals, debugging techniques, and production ML patterns
- Present technical deep-dives to cross-functional stakeholders
Minimum Qualifications
- Education: Bachelor's in Computer Science, Computer Engineering, Electrical Engineering, or related field
- Experience: 3-6 years building and deploying ML systems in production environments
- Computer Vision: Proven track record shipping CV solutions (object detection, segmentation, tracking, or pose estimation) handling real-world data
- Python Proficiency: Strong software engineering skillsclean code, testing (pytest/unittest), packaging, virtual environments, type hints
- Model Deployment: Experience serving models via REST/gRPC APIs with frameworks like FastAPI, Flask, or TorchServe
- Infrastructure: Hands-on with Docker, CI/CD tools (GitHub Actions, GitLab CI), and cloud platforms (AWS/Azure/GCP) or on-prem GPU clusters
- Performance Tuning: Practical experience profiling code (cProfile, py-spy), optimizing memory usage, and reducing inference latency
Preferred Qualifications
- Master's degree in Computer Science, Data Science, Machine Learning, or related field
- Advanced CV: Multi-object tracking (SORT, DeepSORT, ByteTrack), trajectory forecasting, or video understanding models
- Model Serving: Experience with Triton Inference Server, TorchServe, vLLM, or TensorRT optimizations
- LLM/RAG Systems: Built retrieval-augmented generation pipelines using vector databases (Pinecone, Weaviate, Milvus) and embedding models
- Edge Deployment: Optimized models for edge devices (NVIDIA Jetson, Coral TPU) with latency/power constraints
- MLOps Maturity: Worked with experiment tracking (MLflow, Weights & Biases), feature stores (Feast, Tecton), or Kubernetes operators (KubeFlow, Seldon)
- Distributed Training: Experience with multi-GPU training (DDP, DeepSpeed) or large-scale data processing (Ray, Dask)
