About AgZen

AgZen is a fast-growing precision agriculture company headquartered in Somerville, MA, built on MIT research and focused on one problem: making crop spraying more efficient. Our flagship product, RealCoverage, is the world's first system that measures and controls droplet coverage at the leaf level, giving growers real-time visibility into spray performance and cutting chemical and water use by up to 50% without sacrificing yield.

We are a small, technically deep team working at the intersection of fluid mechanics, computer vision, AI, and real agricultural environments. If you want to build technology with measurable impact on how the world grows food, this is the place to do it.

About the Role:

As a Senior Perception Engineer, you will be a key contributor to AgZen’s sensing and software portfolio. This role is at the intersection of agrochemical spraying, machine learning, and computer vision, where you will own and develop the perception platform that helps growers make spraying more efficient. This role will be an essential component of AgZen’s measurement focus group. Strong communication, flexibility, teamwork, the desire to take on different responsibilities and own them will all be essential skills for a successful applicant.

📍This role is located in Somerville, MA (Boston area) with work required to be in-person.

What You’ll Do

Develop advanced computer vision models that support real-time detection and decision-making in AgZen’s spraying systems.

Implement deep-learning models for tasks including object segmentation, classification, optical flow, and 3D reconstruction.

Architect and build new end-to-end models, potentially merging existing networks for improved performance.

Employ techniques like quantization to enable power-efficient model execution on edge devices.

Develop innovative methods to improve model accuracy and inference speed in edge deployment scenarios.

Collaborate with the software engineering team to architect and develop data processing and annotation pipelines for continuous system improvement.

Collaborate closely with hardware, firmware, and product teams to ensure vision models integrate seamlessly with system components.

Analyze field performance to identify root causes of model behavior, propose improvements, and validate updates through experimentation.

Establish best practices for versioning, testing, deployment, and continuous improvement of computer vision models.

Communicate progress, risks, and recommendations to engineering and product leadership in a clear and structured manner.

Stay at the forefront of AI research by exploring and applying advancements in CNNs, transformers, and model quantization.

What We're Looking For

Required:

Bachelor’s or graduate degree in Computer Science, Electrical Engineering, or a closely related field

5+ years of experience in solving problems in computer vision/perception with a foundation in deep learning and ML

Experience working with real-time or edge-based inference and integrating models with hardware or embedded systems.

Demonstrated ability to design experiments, analyze results, and iterate quickly based on field and product requirements.

Deep experience training and validating computer vision models (e.g., object detection, segmentation, and classification) for robust performance in real-world environments.

Proven ability to architect end-to-end computer vision pipelines, strategically decomposing complex problems across multiple models and stages.

Expertise in MLOps best practices and tooling (e.g., MLFlow, W&B, or similar experiment tracking and deployment frameworks) for managing the entire model lifecycle from research to production at scale.

Advanced knowledge of software engineering principles, including software design, source control management, build processes, code reviews, and testing methods

Advanced scientific Python (NumPy, Pandas, scikit-learn) and hands-on experience with deep learning frameworks including PyTorch and/or TensorFlow

Adept with the latest research in perception and machine learning, demonstrated by activity in the research community (e.g., reading conference papers, publications).