At a glance
Design, train, and deploy efficient computer vision models for Deepnight’s ultra-low-light camera, pushing real-time AI inference on embedded hardware for robots, drones, vehicles, and displays.
Summarized by AI from the original posting
What you'll do
- Design novel neural architectures
- Train models
- Deploy real-time inference on embedded hardware
- Advance low-light imaging
- Improve image quality and reduce model latency
- Research computational photography and efficient neural architectures
Requirements
- 01Proficiency with PyTorch
- 02Strong background in computer vision
- 03U.S. citizenship required for the government contract
- 04Master’s degree in computer science or 2 years of work experience
- 05Relevant research publications preferred
Skills
Full description
Deepnight is creating the best low light camera in the world by embedding an AI model into it. Our model processes the camera’s extremely limited signal with an edge AI chip, and can handle light levels as dark as overcast moonless starlight (0.1 millilux), which is as dark as it can possibly be while on the exposed surface of the earth. The camera is one cubic inch, and we have optimized our model for mobile constraints - working at 90 FPS and consuming only 1 watt. We are making all robots, drones, autonomous vehicles, and head-mounted displays able to perform in total darkness.
Deepnight is looking for a computer vision & deep learning engineer to advance the state of the art in low-light imaging. You will design novel neural architectures, train models, and deploy real-time inference on embedded hardware. You’ll utilize your current expertise along with the information gleaned from public journals to push image quality and reduce model latency.
Qualifications
- Proficiency with PyTorch.
- Strong background in computer vision.
- This position supports a U.S. government contract that requires all employees performing the work to be U.S. citizens. Please confirm that you meet this contract requirement before applying.
- Master’s degree in computer science or 2 years of work experience.
Preferred Qualifications
- Experience in training and deploying quantized neural networks.
- Experience in one or more of the following on-device AI tools: QNN, AIMET, TensorRT, SNPE.
- Experience in neural network compression techniques: neural architecture search, knowledge distillation, pruning.
- Experience in low level vision (denoising, super-resolution, frame interpolation).
- Understanding of CMOS imaging (rolling vs. global shutter, read noise, dynamic range, demosaicing, dynamic range, etc.).
- Relevant research publications.
About Deepnight
- Founded in Y-Combinator startup accelerator in January 2024.
- Series A, raised $14M venture funding to date.
- $2M yearly revenue. Several partnerships including with Sony sensors division, US Army & Air Force, MIT
- 15 person highly technical team, growing fast.
About Deepnight
DeepNight is building night vision with AI software. We are pushing the state of the art of computational imaging to build a new generation of night vision devices that see in the dark using AI. We conduct novel research in computational photography, novel efficient neural architectures, and model compression techniques to build extremely efficient AI on the edge.
Compensation: $160K - $240K
Skills: Torch/PyTorch, Machine Learning, Deep Learning, Computer Vision
Experience: 1+ years