Senior AI/ML Engineer – Real-Time Computer Vision & Neural Graphics

Job Summary We are seeking an expert AI/ML Engineer to build photorealistic, zero-latency virtual try-on models (hair styling, hair coloring, and 3D facial augmentation) that execute entirely on a live camera stream. You will be responsible for creating hyper-accurate sub-pixel hair segmentation, real-time alpha matte prediction, physics-aware lighting shaders, and low-latency neural rendering pipelines that operate at 30–60 FPS on iOS, Android, and Web platforms. Key Responsibilities Real-Time Live-Camera Segmentation: Develop lightweight neural networks for real-time fine-strand hair segmentation, semantic edge matting, and face mesh tracking directly on live video feeds. Photorealistic Lighting & Shaders: Build custom custom WebGL/Metal/Vulkan graphics shaders and neural recoloring pipelines that preserve natural specular highlights, shadows, transparency, ambient lighting, and dark-to-light hair dye blending. 3D Mesh Alignment & Physics: Implement real-time 3D head-tracking (3DMM) to dynamically anchor 3D hair models to head movements with realistic movement, occlusion, and depth sorting. Edge Optimization (Mobile & Web): Quantize, prune, and optimize deep learning models (TFLite, CoreML, WebGPU/ONNX) to achieve 60 FPS performance with zero latency and low battery/thermal consumption on live mobile cameras. Synthetic & Real-World Dataset Pipeline: Create automated data collection and synthesis pipelines to train models against diverse lighting conditions, hair textures, complex backgrounds, and motion blur. Required Qualifications & Skills Core ML & Vision: 4+ years of experience in PyTorch/TensorFlow, with specialized expertise in Semantic Segmentation, Alpha Matting (e.g., Deep Image Matting), and Real-Time Video Processing. Neural Rendering & Shaders: Strong understanding of graphics pipelines, custom GLSL/HLSL shaders, WebGL/WebGPU, and rendering engine integrations (Three.js, Unity, or native Metal/Vulkan). On-Device Edge Deployment: Proven track record of deploying models to edge devices using CoreML (iOS), TFLite / NNAPI (Android), and ONNX / WebNN (Web) running at sub-30ms per frame. Face & Head Tracking: Practical experience with facial keypoint algorithms, MediaPipe, ARKit/ARCore mesh tracking, and spatial registration. High-Performance Code: Proficiency in C++, Python, and Swift/Kotlin for writing low-latency native camera frame buffer hooks. Preferred Qualifications Experience working on commercial AR beauty SDKs (e.g., ModiFace, Banuba, DeepAR). Familiarity with neural radiance fields (NeRFs), 3D Gaussian Splatting, or Neural Physics for real-time strand animation.