Machine Learning Engineer

Role Overview

You will design, optimize, and deploy real-time computer vision and machine learning pipelines for a multi-sensor biometric authentication device. This includes processing high-resolution data from RGB, IR, and depth/thermal sensors, implementing robust anti-spoofing mechanisms, and integrating privacy-preserving encrypted face-matching algorithms (e.g., CKKS/FHE). The role demands expertise in embedded GPU platforms (NVIDIA Jetson series) and the ability to push models to production under strict latency and security constraints.

Responsibilities

Model Development & Optimization

• Build and fine-tune face detection, alignment, and embedding extraction models for RGB + IR + depth + thermal data.
• Research and implement spoof detection (e.g., texture analysis, depth cues, challenge-response).
• Quantize and optimize models for Jetson hardware (TensorRT, ONNX Runtime, CUDA/CuDNN).

Encrypted Matching Pipeline

• Implement vector encryption (CKKS/FHE) for face embeddings.
• Develop threshold-based similarity checks in encrypted space (GPU-accelerated).
• Collaborate on algorithm selection (cosine similarity vs. alternatives) ensuring high precision at 95%+ thresholds.

Multi-Sensor Data Fusion

• Synchronize and process input from multiple camera modules.
• Fuse thermal/IR/depth data to improve accuracy and spoof resistance.

Real-Time Performance Engineering

• Achieve sub-second processing latency.
• Optimize compute graph, memory usage, and sensor I/O.

Hardware Integration

• Work closely with embedded engineers to integrate ML pipelines into the device’s OS and BSP.

R&D

• Stay updated on state-of-the-art in biometric security, encrypted ML, and edge AI.
• Prototype and test new methods for privacy-preserving identity verification.

Required Skills

Core ML / CV Skills

• Strong background in computer vision (OpenCV, PyTorch/TensorFlow).
• Experience with face recognition systems (e.g., ArcFace, FaceNet) and anti-spoofing.
• Model optimization for constrained devices (TensorRT, pruning, quantization).

Embedded AI

• NVIDIA Jetson platform experience (Nano, Xavier, Orin).
• CUDA, cuDNN, GPU profiling, and performance tuning.

Privacy-Preserving ML

• Hands-on with Fully Homomorphic Encryption (CKKS) / SMPC or related frameworks (e.g., OpenFHE, SEAL, FIDESlib).
• Understanding of secure enclaves and encrypted inference.

Systems & Integration

• Comfortable working with BSP bring-up, camera sensor integration, and custom drivers.
• Strong Python & C++ skills for production deployment.

Bonus

• Depth/thermal imaging experience.
• Knowledge of biometric standards & certification (ISO/IEC 19794-5).
• Background in security for identity systems.

Ideal Candidate

• Has deployed ML models on embedded GPU hardware.
• Has worked on biometric authentication or high-security identity verification projects.
• Can own the full stack of data processing — from raw sensor input to encrypted decision output.
• Thrives in R&D-heavy, rapid-prototyping environments.