Senior AI/ML Engineer (Technical Lead)

Softengi is seeking a highly experienced Senior AI/ML Engineer to lead the creation of a next-generation real-time biometric estimation system for global sports broadcasts. The role focuses on building a non-contact system using high-speed video and computer vision to estimate athlete exertion with 85%+ accuracy for live sports applications.

• Phase 0 - Pilot: Design ML architecture for grip estimation, train initial models using force sensors and video data, build training and data labeling pipeline, validate across diverse athlete profiles
• Phase 1 - Demo: Refine models using demo event data, implement 2-second per-athlete calibration, optimize for edge cases including skin tones and hand sizes
• Phase 2 - Production: Improve accuracy toward 85%+, optimize inference to <40ms (P99), implement advanced scoring features, and handle production hardening and reliability improvements
• Make critical architectural decisions and lead technical strategy
• Systematically experiment to achieve high model accuracy requirements

• 5+ years of experience in applied machine learning/deep learning
• Strong background in computer vision and image-based regression tasks
• Hands-on experience with PyTorch or TensorFlow for production systems
• Experience with real-time inference optimization (TensorRT, ONNX Runtime)
• Understanding of signal processing and temporal data analysis
• Experience training models on diverse datasets with bias handling and data augmentation
• Proven track record achieving high-accuracy requirements (>80% on complex tasks)
• Strongly preferred: biomechanical/physiological signal estimation, optical phenomena knowledge, few-shot learning, transfer learning, sports analytics background
• Nice to have: NIR imaging, multi-spectral analysis, pose estimation frameworks, NVIDIA edge ML deployment, time-series forecasting

Technical leadership role with responsibility for critical architectural decisions under uncertainty. Must demonstrate problem-solving abilities for novel ML challenges, clear communication of complex concepts to non-technical stakeholders, adaptability to real-world performance feedback, and full ownership of model accuracy through systematic experimentation.