Lead AI Researcher

The role

A hands-on Lead AI Researcher will set the technical roadmap, review architecture, and manage a small high-performing AI team. The role involves shipping AI/ML models from prototype to production and applying LLMs or multimodal models to video applications.

What you will do

• Set the technical roadmap
• Review architecture
• Mentor engineers
• Manage a small high-performing AI team
• Ship AI/ML models from prototype to production at scale
• Perform model post-training including fine-tuning, preference optimization, evaluation, and learning from user feedback
• Build data flywheels that turn user behavior and feedback into training data
• Optimize inference for self-hosted models including latency, throughput, GPU utilization, quantization, and serving cost
• Apply LLMs or multimodal models to video applications
• Make build-vs-buy decisions across proprietary APIs, open-source models, and internally trained models

Required qualifications

• Proven experience shipping AI/ML models from prototype to production at scale
• Deep expertise in model post-training, including fine-tuning, preference optimization, evaluation, and learning from user feedback
• Experience building data flywheels that turn user behavior and feedback into training data and continuous model improvements
• Strong inference optimization experience for self-hosted models, including latency, throughput, GPU utilization, quantization, and serving cost
• Strong model evaluation skills, including offline benchmarks, human evaluation, online experiments, and product-quality metrics
• Practical experience applying LLMs or multimodal models to video applications, such as highlight detection, content curation, ranking, personalization, and editing decisions
• Familiarity with video-related models and systems, including but not limited to vision-language models, ASR/transcription, video understanding, and enhancement/upscaling
• Able to make build-vs-buy decisions across proprietary APIs, open-source models, and internally trained models
• Capable of setting the technical roadmap, reviewing architecture, mentoring engineers, and managing a small high-performing AI team
• Product-oriented and pragmatic

Nice to have

• People management
• Experience with consumer video, creator tools, recommendation/content systems, or other high-scale multimodal products
• Distributed training infrastructure and inference optimization experience for self-hosted models: including latency, throughput, GPU utilization, quantization, and serving cost
• Familiarity with video-related models and systems: vision-language models, ASR/transcription, video understanding/reasoning/generation, and enhancement/upscaling

About the company

OpusClip is an AI video agent company headquartered in Mountain View with a team of 100 enthusiasts and video experts. They have raised $50 million in total funding from investors including SoftBank Vision Fund and Samsung Next.