Machine Learning Engineer II, Computer Vision

Amniscient LLC•Houston, TX

17h•$100,000 - $150,000

About The Position

Amniscient is an innovative start-up developing machine learning and computer vision technology for the retail industry. The Machine Learning Engineer II will build and maintain platform capabilities that enable customers to create their own computer vision models. The platform turns raw images into labeled datasets, trains models at scale on GPU clusters, benchmarks models, and supports deployment to production. The role owns features from day one and collaborates with data, platform, and product teams, with support from senior engineers. The engineer will train, fine-tune, evaluate, optimize, deploy, and monitor computer vision models, particularly object detectors. The position also involves building data pipelines and labeling workflows, defining evaluation methods, running reproducible experiments, writing tested Python code, participating in code reviews, and explaining findings to product and engineering stakeholders.

Requirements

2–5 years of professional experience building ML systems, including at least 2 years focused on computer vision.
Bachelor of Science in Computer Science, Computer or Electrical Engineering, or a related field.
Have shipped at least one object detection model to production and maintained it after launch.
Strong Python skills and hands-on experience with PyTorch.
Working knowledge of modern one-stage, two-stage, and transformer-based detection architectures.
Understanding of detection fundamentals, including anchors and anchor-free methods, IoU, Non-Maximum Suppression, mAP, and class imbalance.
Experience with image augmentation, dataset curation, and annotation quality control.
Comfortable with Git, Linux, Docker, and training on GPUs.

Nice To Haves

MS or PhD with a Computer Vision focus.
Experience with multi-object tracking, instance segmentation, or video-based detection.
MLOps experience, including experiment tracking, model registries, CI/CD for models, and production monitoring.
Experience with distributed training on multiple GPUs.
Experience with edge deployment (NVIDIA Jetson, mobile, or embedded) and inference optimization with ONNX Runtime or TensorRT.
Publications, open-source contributions, or competition results in computer vision.

Responsibilities

Train, fine-tune, and evaluate object detection models, such as YOLO-family, DETR-family, and Faster R-CNN, for production use cases.
Build and maintain image and video data pipelines, including labeling workflows and quality checks.
Define evaluation metrics, including mAP, per-class precision and recall, latency, and IoU, and perform error analysis to determine improvements.
Optimize models for deployment using quantization, pruning, and export to ONNX or TensorRT.
Deploy and monitor production models, tracking accuracy drift and data distribution shift.
Run experiments with clear tracking and reproducible results.
Write clean, tested Python code and participate in code reviews.
Share findings with product and engineering stakeholders in plain terms.
Write and maintain platform features that enable users to build their own models.