AI/ML Engineer

CACI•Ashburn, VA

8h•Hybrid

About The Position

CACI is seeking a Mid-Level AI/ML Computer Vision Engineer to join their BEAGLE Agile Solution Factory (ASF) Team, supporting the Customs and Border Protection (CBP) client in Northern Virginia. The role involves working within a specialized 3-person pod focused on building custom object detection and classification models using the YOLO framework. The primary mission is to process live and recorded Full-Motion Video (FMV) feeds to enable automated parsing of entities such as roads, vehicles, people, and terrain features. This position requires hands-on involvement in bringing raw production video into a secure training environment, curating datasets, training models, and deploying high-performance inference pipelines.

Requirements

College degree (B.S.) in Computer Science, Data Science, Software Engineering, or a related discipline.
5 or more years of software engineering or data science experience, with 2+ years of hands-on experience building/training Computer Vision models.
Programming in Python and using core machine learning libraries (NumPy, Pandas, Scikit-Learn).
Hands-on experience with PyTorch or TensorFlow for training neural networks.
Experience implementing YOLO (e.g., YOLOv8/v9) for object detection, segmentation, or classification.
Familiarity with image and video processing techniques using OpenCV or FFmpeg.
Data curation, formatting, and experience using data annotation platforms (e.g., CVAT, LabelImg).
Using Git for version control and collaborating on shared codebases.
Strong analytical skills, with a focus on benchmarking model performance (mAP, FPS, inference latency).
Willing to work on site 2-3 days a week.
Pass CBP background investigation (U.S. Citizenship required). Criteria include, but are not limited to: 3-year check for felony convictions, 1-year check for illegal drug use, 1-year check for misconduct such as theft or fraud.

Nice To Haves

Familiarity with containerized development environments (Docker).
Experience with GPU execution environments and model export formats (ONNX, TensorRT).
Experience with multi-object tracking libraries.
Familiarity with Agile tools like Jira and Confluence.

Responsibilities

Train, fine-tune, and evaluate custom YOLO object-detection models on proprietary datasets.
Run experiments with different hyperparameters, augmentations, and model sizes to find the optimal balance of speed and accuracy.
Process, clean, and format raw video files recorded from production environments into high-quality datasets for training.
Manage the data annotation lifecycle, labeling images and video frames, verifying annotation quality, and generating synthetic data where needed.
Write Python scripts to parse video frames, feed them into inference engines, and export structured prediction data (JSON/Protobuf) representing detected objects.
Generate performance reports, confusion matrices, and precision-recall curves to identify model blind spots and continuously improve detection capabilities.
Work closely with the Senior AI/ML Lead, video engineers, and developers in an Agile/Scrum framework to deliver production-ready code in short sprints.