Chooch
Website:
chooch.com
Job details:
Company Description Chooch AI is a computer vision platform that helps hospitals automate supply room monitoring and inventory management. Using fixed cameras and AI, the platform continuously tracks supplies in storage areas and updates on-hand counts in real time. When inventory levels fall below defined thresholds, replenishment orders are automatically generated and sent to ERP systems, eliminating manual counts and barcode scanning. Hospitals using Chooch AI significantly reduce critical stockouts, optimize inventory levels, and improve supply chain staff productivity. By turning supply rooms into self-managing systems, Chooch AI enables clinical and supply chain teams to focus more time on patient care.
Responsibilities/What you'll do at Chooch:
- Design, train and fine-tune models for object detection, classification, segmentation, tracking and anomaly detection on real customer video and images.
- Adapt pretrained and foundation models (YOLO-family detectors, ViTs, CLIP-style and vision-language models) to new customer use cases with limited labeled data.
- Own data pipelines end to end: curation, labeling guidelines, augmentation, synthetic data, and dataset versioning.
- Optimize models for production and edge hardware using quantization, pruning, distillation, ONNX export and TensorRT, hitting agreed latency and accuracy targets.
- Build evaluation suites with clear metrics (mAP, precision/recall, false-alarm rates per camera-hour) and catch regressions before release.
- Deploy and monitor models in production, diagnose drift and failure cases, and close the loop with retraining.
- Work with product, solutions engineers and customers to turn vague requirements into measurable model specs.
- Write clean, tested, reviewed code and contribute to shared ML tooling.
Required qualifications (must have)
- 3–5 years of industry experience building ML systems, with at least 2 years focused on computer vision.
- BS or MS in Computer Science, Electrical Engineering, Math or a related field, or equivalent practical experience.
- Strong Python and hands-on experience with PyTorch (or TensorFlow) for training and debugging deep learning models.
- Shipped at least one computer vision model to production, and can explain the trade-offs made along the way.
- Solid grasp of CNN and transformer architectures for detection, segmentation and classification.
- Experience with OpenCV and image/video processing: decoding streams, preprocessing, augmentation.
- Working knowledge of model optimization for inference (ONNX, TensorRT, quantization or similar).
- Comfortable with Docker, Git, Linux and training on GPUs in a cloud environment (AWS/Azure).
- Sound ML fundamentals: loss functions, overfitting, class imbalance, evaluation metrics and experiment design.
- Clear written and verbal communication, including explaining model behavior to non-ML stakeholders.
Nice to have
- Experience with vision-language models (CLIP, LLaVA-style models) or open-vocabulary detection.
- Multi-object tracking and video analytics (re-identification, action recognition, temporal models).
- Few-shot learning, active learning, or synthetic data generation for low-data domains.
- MLOps tooling: MLflow or Weights & Biases, Kubernetes, Triton Inference Server, CI/CD for models.
- Domain experience in Healthcare, retail and supply chain, manufacturing inspection, workplace safety or video surveillance.
Click on Apply to know more.
