Senior Computer Vision Engineer

Diffraqtion; Somerville (MA); On-site; USD 140,000 - 190,000; Full time

Benefits offered by this job

Medical, dental and vision on 1st day

Company 401K match

Stock options

Job summary

Diffraqtion is building a Quantum Vision System; you will join early to design and deploy computer-vision models that run on hardware destined for space. You will own the imaging pipeline from raw sensor data to actionable detections, perform robust de-noising and reconstruction, and collaborate with optical/embedded engineers to maximize end-to-end performance.

The role requires a MS and 3+ years in CV or perception on real sensors, with strong Python/PyTorch skills and C++ knowledge.

Qualifications

• MS in CS, EE, or related field.
• 3+ years of industry experience in CV or perception on real sensors.
• Hands-on multi-sensor fusion using probabilistic/learned methods.
• Hands-on computational imaging: denoising/restoration.
• Strong grounding in estimation theory and inverse problems; DL for imaging.
• Experience designing robust experiments with baselines and uncertainty quantification.
• Proficiency in Python and PyTorch; working knowledge of C++.
• Solid understanding of image formation, camera models, calibration, and sensor noise.
• Solid software engineering practices: version control, testing, reproducible experiments.
• On-site in Somerville, MA with hardware team.

Responsibilities

• Design, train, and deploy CV models for imaging and perception tasks like reconstruction, detection, classification, and tracking tailored to our optical front end.
• Develop de-noising and image-restoration methods robust to sensor noise, radiation, low light, and aberrations.
• Build end-to-end model of the imaging chain coupling physics-based optical and sensor models with neural networks for joint optics/algorithm optimization.
• Design sensor fusion algorithms that combine camera data with other onboard sensors using probabilistic and learned approaches.
• Integrate vision and fusion outputs into the real-time control loop with defined interfaces and latency budgets.
• Optimize and deploy models to edge hardware (embedded GPU, FPGA, SoC) with quantization and pruning.
• Build data pipelines and simulation tools for synthetic and lab datasets, including calibration and domain-gap analysis.
• Define metrics and run hardware-in-the-loop experiments to validate performance on real sensor data.
• Document designs and contribute to reviews and the codebase.

Skills

Python

PyTorch

C++

Estimation theory

Inverse problems

Multi-sensor fusion

Probabilistic methods

Education

M.S. in Computer Science, Electrical Engineering, or related field

Tools

TensorRT

ONNX Runtime

NVIDIA Jetson

Job description

About Diffraqtion
Diffraqtion builds a novel Quantum Vision System that pulls more information out of each photon than conventional imaging allows. Our photonics process incoming light before it reaches the sensor, so the computation happens passively in the light itself. We are an early-stage team in Somerville by founders from MIT and Harvard, working under NASA and DARPA contracts and backed by Lockheed Martin and Sumitomo. Our systems are headed for satellites that protect assets in orbit and on the ground, and for telescopes hunting Earth‑like planets around other stars. You would be joining early, with real hardware to build and real missions to put it on.

About Diffraqtion
Diffraqtion builds a novel Quantum Vision System that pulls more information out of each photon than conventional imaging allows. Our photonics process incoming light before it reaches the sensor, so the computation happens passively in the light itself. We are an early-stage team in Somerville by founders from MIT and Harvard, working under NASA and DARPA contracts and backed by Lockheed Martin and Sumitomo. Our systems are headed for satellites that protect assets in orbit and on the ground, and for telescopes hunting Earth‑like planets around other stars. You would be joining early, with real hardware to build and real missions to put it on.

The Role

We are hiring a Computer Vision Engineer to build the models at the center of our Quantum Vision System. Our photonics do part of the computation in light before it reaches the sensor and your models finish it. You will own the imaging pipeline from raw sensor output to actionable detections: physics‑based reconstruction and de‑noising, detection and tracking, and deployment on the embedded hardware that flies. You will work directly with our optical and opto‑mechanical engineers to build an end‑to‑end model of the imaging chain, from photons to outputs, so optics and algorithms can be optimized together. The first systems you work on are headed to space.

You have built multi‑sensor fusion on real sensors and have shipped estimators and learned models that had to work on hardware, not just on benchmarks. You are comfortable with probabilistic models and deep learning in the same week. You want a problem nobody has solved yet and the hardware to test it on.

Note: This position requires U.S. citizenship, lawful permanent resident status, or eligibility for authorization under U.S. export control regulations (ITAR/EAR).

Key Responsibilities

• Design, train, and deploy computer vision models for imaging and perception tasks like image reconstruction, detection, classification, and tracking tailored to our optical front end.
• Develop de‑noising and image‑restoration methods (classical and learned) that are robust to sensor noise, radiation effects, low light, and optical aberrations.
• Build a cohesive, end‑to‑end model of the imaging chain coupling physics‑based optical and sensor models with neural networks to enable joint optimization of optics and algorithms.
• Design and implement sensor fusion algorithms that combine camera data with other onboard sensors using probabilistic and learned approaches.
• Integrate vision and fusion outputs into the system's real‑time control loop, defining interfaces, latency budgets, and failure modes together with controls and embedded engineers.
• Optimize and deploy models to resource‑constrained edge hardware (e.g., embedded GPU, FPGA, or SoC), including quantization, pruning, and runtime profiling.
• Build data pipelines and simulation tools for synthetic and lab‑captured datasets, including calibration, labeling, and domain‑gap analysis between simulation and hardware.
• Define metrics and run hardware‑in‑the‑loop experiments in the lab to validate algorithm performance on real sensor data.
• Document designs and results clearly, and contribute to technical reviews and the team's shared codebase and best practices.

Qualifications

Required:

• M.S. in Computer Science, Electrical Engineering, or a related engineering field.
• 3+ years of industry experience developing and deploying computer vision or perception systems on real sensors.
• Hands‑on experience with multi‑sensor measurement fusion: combining measurements from multiple imaging or onboard sensors into one estimate or image, using probabilistic and learned methods.
• Hands‑on experience with computational imaging: image de‑noising, restoration, or reconstruction.
• Strong grounding in estimation theory and inverse problems and hands‑on experience with deep learning for imaging.
• Experience designing experiments that hold up: ablations, baselines, uncertainty quantification.
• Strong proficiency in Python and PyTorch (or equivalent deep learning framework), with working knowledge of C++.
• Solid foundation in image formation, camera models, and calibration, and in the noise characteristics of imaging sensors.
• Solid software engineering practices: version control, testing, and reproducible experiments.
• This an on‑site in Somerville, MA, alongside the hardware team.

Preferred

• Background in aerospace, defense, robotics, or autonomous systems, ideally with space or airborne sensors.
• Robotics state estimation: Kalman filters, factor graphs, fusing IMU, GPS, and lidar.
• Physics‑informed or differentiable optics models, or end‑to‑end optics‑algorithm co‑design.
• Deploying models to edge or embedded platforms (e.g., TensorRT, ONNX Runtime, NVIDIA Jetson).
• FPGA or hardware‑accelerated inference (e.g., Vitis AI, HLS).
• Non‑visible imaging (e.g., infrared, multispectral, or hyperspectral) or event‑based sensors.
• Publications or open‑source contributions in sensor fusion, computational imaging, or computer vision.

Benefits

• Competitive salary and stock options.
• Medical, dental and vision on 1st day.
• Company 401K match.
• Ownership in the company and seeing your code fly in space.

Equal Opportunity Employer

Diffraqtion is committed to fostering an inclusive and meritocratic workplace. We provide equal employment opportunity to all employees and applicants without regard to race, color, religion, sex (including pregnancy), sexual orientation, gender identity, national origin, ancestry, citizenship status, age, disability, genetic information, military or veteran status, or any other status protected by federal, state, or local law. Diffraqtion is committed to equal opportunity for protected veterans and individuals with disabilities. Applicants who need a reasonable accommodation during the application or interview process may contact [e-mail].

IT IS UNLAWFUL IN MASSACHUSETTS TO REQUIRE OR ADMINISTER A LIE DETECTOR TEST AS A CONDITION OF EMPLOYMENT OR CONTINUED EMPLOYMENT. AN EMPLOYER WHO VIOLATES THIS LAW SHALL BE SUBJECT TO CRIMINAL PENALTIES AND CIVIL LIABILITY.