Machine Learning Engineer

PRADCO Inc.; Massachusetts; On-site; USD 120,000 - 150,000; Full time

Job summary

PRADCO Inc. is hiring a Machine Learning Engineer in Massachusetts for an onsite role. You will design and train object detection models, build end-to-end ML pipelines, and develop deer re-identification methods using coat patterns. Collaborate with data engineers and a wildlife biologist to validate outputs.

You will deploy on GPU infrastructure (AWS SageMaker, GCP Vertex AI), using Mlflow or Weights & Biases for experiment tracking and retraining to handle data drift across seasons.

Qualifications

• 4+ years of experience in machine learning engineering with production deployments.
• Strong PyTorch proficiency, with experience in Ultralytics/YOLO or similar detection frameworks.
• Deep understanding of CNN architectures, transfer learning, and domain adaptation.
• Experience deploying models at scale on GPU infrastructure (AWS SageMaker, GCP Vertex AI).
• Proficiency in Python and data pipeline tooling like Kafka and Airflow.
• Solid ML evaluation skills (confusion matrices, mAP, precision/recall) and diagnosing failures.
• Familiarity with time-series methods (LSTMs, Prophet, XGBoost) for temporal data.

Responsibilities

• Design and train object detection/classification models for wildlife imagery.
• Build end-to-end ML pipeline: data ingestion, preprocessing, training, eval, deployment.
• Develop deer re-identification models using coat patterns and antler morphology.
• Create feature engineering by combining vision outputs with environmental signals.
• Establish ML Ops practices with Mlflow or Weights & Biases for experiment tracking.
• Collaborate with Data Engineering and Wildlife Biologist advisor; monitor production performance.

Tools

PyTorch

Ultralytics/YOLO

SageMaker

Vertex AI

Kafka

Airflow

Mlflow

Weights & Biases

Triton

TorchServe

AWS SageMaker

GCP Vertex AI

Python

CUDA/GPU

Job description

PRADCO Inc. is hiring a Machine Learning Engineer in Massachusetts for an onsite role. The position supports the full prediction machine learning lifecycle, from tagged trail camera imagery through deer movement predictions and hunt location optimization, enabling stand recommendations based on behavioral signals.

Responsibilities

• Design and train object detection and classification models (such as YOLOv8, RT-DETR, or similar) to detect deer presence along with sex, age class, and antler characteristics in trail camera images.
• Build and maintain an end-to-end ML pipeline, including data ingestion from cloud storage, preprocessing, model training on GPU clusters, evaluation, and deployment using Triton, TorchServe, or equivalent tools.
• Develop deer re-identification models that use coat patterns and antler morphology to track individual animals across cameras and time.
• Engineer features by combining vision outputs with environmental signals such as weather, terrain, moon phase, and a rut calendar to support downstream behavioral prediction models.
• Apply ML Ops practices using Mlflow or Weights & Biases for experiment tracking, model versioning, and staged production deployments.
• Work with Data Engineering to optimize data pipelines and collaborate with a Wildlife Biologist advisor to validate model outputs against real-world deer behavior.
• Monitor production performance and implement retraining pipelines to mitigate data drift across seasons.

Requirements

• 4+ years of experience in machine learning engineering, including demonstrated production deployments.
• PyTorch proficiency, with strong preference for experience using Ultralytics/YOLO or comparable detection frameworks.
• Deep understanding of CNN architectures, transfer learning, and domain adaptation.
• Experience deploying models at scale on GPU infrastructure, including AWS SageMaker, GCP Vertex AI, or equivalent platforms.
• Proficiency in Python and familiarity with data pipeline tooling such as Kafka and Airflow (or similar).
• Strong ML evaluation fundamentals, including confusion matrices, mAP, precision/recall tradeoffs, and the ability to diagnose model failures.
• Familiarity with time-series prediction methods such as LSTMs, Prophet, and XGBoost for temporal data.

Essential Job Function

• Experience with re-identification (RelD) or few-shot learning tasks.
• Prior work involving wildlife imagery, agricultural computer vision, or other domains characterized by low contrast and occlusions.
• Experience with Microsoft Azure.
• Genuine interest in the outdoors or hunting, with domain empathy considered a positive factor for product quality.

Technologies

• PyTorch, Ultralytics/YOLO, YOLOv8, RT-DETR
• Triton, TorchServe
• Mlflow, Weights & Biases
• AWS SageMaker, GCP Vertex AI
• Python, Kafka, Airflow
• CNN architectures, LSTMs, Prophet, XGBoost

Location and Experience

• Location: Massachusetts (onsite)
• Minimum experience: 4 years