Responsibilities
- Design, train, and iterate on custom object detection models specifically tuned for retail environments, inventory tracking, and product recognition
- Fine-tune and deploy open-source vision-language models (LLaVA, Qwen-VL, InternVL, PaliGemma, etc.) for product understanding, zero-shot classification, and scene reasoning; build vision-language-action pipelines that translate visual understanding into downstream decisions
- Take state-of-the-art models and make them blazingly fast for edge deployment through quantization, pruning, and architectural optimization
- Build robust data pipelines and annotation workflows to continuously improve model performance on diverse retail scenarios
- Stay ahead of the curve on CV and VLM research, prototype new architectures, and determine what's actually production-ready versus academic noise
- Mentor engineers, establish best practices for model development, and drive technical decisions around our CV infrastructure
Requirements
- 4+ years of hands-on computer vision engineering, with a proven track record of shipping models to production
- Deep expertise with YOLO and YOLO-E architectures - you've trained them, tuned them, and know their quirks intimately
- Hands-on experience with open-source VLMs (LLaVA, Qwen-VL, InternVL, PaliGemma, or similar) - fine-tuning, evaluation, and production deployment
- Familiarity with VLA frameworks and applying vision-language-action models to real-world perception and decision tasks
- Edge deployment mastery - experience with TensorRT, ONNX Runtime, or similar frameworks for optimizing models for constrained devices, including quantized VLMs
- Strong software engineering fundamentals - clean code, version control, CI/CD for ML, and the ability to build maintainable systems
- Production ML experience - you understand the difference between a Jupyter notebook and a production-grade ML system
Nice to Have
- Experience developing solutions deployed to the NVIDIA Jetson family of products
- Experience with retail, inventory management, or similar product-focused CV applications
- Background with PyTorch and modern training frameworks (Transformers, LitGPT, Unsloth, etc.)
- Experience running VLM inference efficiently (vLLM, llama.cpp, SGLang, or similar)
- Familiarity with synthetic data generation and data augmentation techniques
- Knowledge of model versioning and experiment tracking (MLflow, Weights & Biases, etc.)
- Publications or open-source contributions in computer vision or multimodal AI
- Experience with AWS: EC2, ECS, Fargate, S3, Bedrock, SageMaker, etc.
