Production CV Engineer for Watermarking & Media Provenance

STEG AI Corporation; Irvine (CA); Hybrid; USD 120,000 - 180,000; Full time

Job summary

Steg.AI in Irvine, CA is seeking a Machine Learning Engineer to productionize steganography models for our watermarking SDKs and cloud platform. You will build scaffolding around models to boost performance and extend support to new image and video formats.

Requirements include 3+ years in computer vision, strong Python/OpenCV/NumPy/ONNX skills, and experience with image/video processing and codecs. Bonus: C/C++, PyTorch, and MLOps tooling.

Qualifications

• 3+ years of full-time industry experience in computer vision or image/video processing.
• Python, OpenCV, NumPy, ONNX, ONNX runtime
• Traditional 2D/3D computer vision techniques: feature matching, object detection, model estimation, optimization, etc.
• Image/signal processing: imaging, compression, color spaces, filtering, resampling, etc.
• FFmpeg and video codecs such as H.265 and HEVC

Responsibilities

• Productionize models with inference pipelines for cloud platform and SDKs.
• Extend watermarking to new file types and codecs.
• Build image, video, and audio pipelines (decode, color conversion, resampling, re-encode).
• Recover watermarks from transformed media using image processing techniques.
• Benchmark quality and robustness, and track regressions.

Skills

Python

OpenCV

NumPy

ONNX

ONNX runtime

Image/video processing

Feature matching

Object detection

Model estimation

Image/signal processing

Tools

FFmpeg

Video codecs (H.265/HEVC)

Pillow

PyTorch

Job description

Steg.AI in Irvine, CA is seeking a Machine Learning Engineer to productionize steganography models for our watermarking SDKs and cloud platform. You will build scaffolding around models to boost performance and extend support to new image and video formats.

Requirements include 3+ years in computer vision, strong Python/OpenCV/NumPy/ONNX skills, and experience with image/video processing and codecs. Bonus: C/C++, PyTorch, and MLOps tooling.