AI Engineer Computer Vision & Physiological Signal Analysis

Role Summary

AI Engineer

Key Responsibilities

• Research, prototype, and productionize

  computer vision models

  for:
  

• Facial expression recognition

  (emotion classification)
  

• Remote photoplethysmography (rPPG)

  for

  BPM

  estimation
  

• SpO estimation

  from RGB video signals
  
• Preprocess webcam images for signal enhancement (ROI extraction, denoising)
  
• Integrate ML models into the backend (FastAPI-based), processing real-time webcam input
  
• Optimize models for latency, accuracy, and resource usage (on CPU/GPU)
  
• Deploy AI models as

  Dockerized microservices

  (REST/WS-based inference APIs)
  
• Work with frontend and backend teams to ensure seamless model integration
  
• Conduct internal validation, A/B testing, and ongoing calibration
  

Tech Stack / Tools

• Languages

  : Python (must), PyTorch or TensorFlow
  

• CV/ML Libraries

  : OpenCV, Mediapipe, dlib, scikit-learn
  

• Signal Processing

  : numpy/scipy, rPPG libraries (e.g., DeepPhys, MTTS-CAN)
  

• Deployment

  : FastAPI, Docker, ONNX, TorchScript
  

• Infra

  : Redis, PostgreSQL, AWS (GPU EC2, S3, ECS)
  

Required Skills

• 3+ years of experience building

  computer vision

  or

  bio-signal AI models

  
  
• Hands-on experience with

  facial emotion recognition

  using CNNs/RNNs/transformers
  
• Familiarity with

  rPPG-based BPM & SpO estimation

  from webcam videos
  
• Strong background in

  signal processing

  , noise reduction, and ROI selection
  
• Ability to write production-grade Python code and deploy AI inference services
  
• Experience optimizing models for inference (quantization, pruning, ONNX)
  

Nice to Have

• Experience with

  real-time WebRTC or webcam processing

  
  
• Familiarity with

  ML pipelines

  (MLflow, ClearML, SageMaker)
  
• Experience with

  face tracking

  (MediaPipe FaceMesh, dlib landmarks)
  
• Familiarity with

  multi-modal emotion detection

  (face + voice)
  
• Knowledge of privacy-preserving AI methods (differential privacy, edge inference)
  

What Youll Own

• Accurate real-time facial emotion AI module
  
• Vital signs (SpO/BPM) module with acceptable medical-grade error margins
  
• Robust backend integration for WebSocket/webcam flow
  
• Scalable AI deployment strategy for 10K 1M users