Higgsfield AI is the fastest-growing generative AI company in history: $500M annual revenue (ARR), 25M+ users worldwide, 6M+ generations per day, and among our clients are 390 Fortune 500 brands. We operate at the very frontier of AI video and next-generation creative tools. Joining Higgsfield means becoming part of a team that is shaping the future of AI-native products, in a company that not only moves fast but redefines the very concept of 'fast'.
What You'll Do
- Analyze motion and physical dynamics in video: optical flow, camera motion vs. object motion, temporal consistency, and physical plausibility — dynamics, collisions, gravity, deformation.
- Train models that will be used in the post-training stage, including reward models.
- Build video classifiers by motion types, frame types, genres, and quality — using both classical methods and VLM/embeddings.
- Develop methods for creating multi-shot video sequences: shot boundary detection, assembling coherent multi-shot scenes with consistent characters and locations, generating descriptions for individual shots.
- Work with audio: speech/music/sound event detection, audio-visual synchronization assessment (lip-sync and sound-to-action correspondence), audio quality filtering, audio track descriptions.
- Apply classical and trainable data curation methods: near-duplicate detection, heuristic filtering, VLM-based captioning, synthetic data generation, dataset versioning.
- Build and maintain large-scale distributed data processing pipelines.
Requirements
- Strong PyTorch skills and practical experience training models, including fine-tuning and deploying classifiers, embedding models, and VLMs.
- Deep knowledge of video processing methods: motion analysis (optical flow, tracking, motion estimation), shot boundary detection, temporal consistency and quality assessment.
- Experience building and validating classifiers on video data: annotation pipelines, active learning, threshold calibration, measuring precision/recall at scale.
- Good understanding of data curation methods: deduplication, filtering, sampling, dataset balancing, and how these decisions impact model training.
- Experience processing large-scale unstructured data, especially video, audio, and images.
What We Offer
- Competitive base salary in USD — depending on your experience, skills, and role level.
- Stock options: participation in the company's stock option program — an opportunity to share in Higgsfield's long-term growth.
- Relocation support to Almaty for candidates moving from another city or country.
- A dynamic environment with a high level of collaboration, where you work directly with experienced leaders and have a real impact on the product and company.
- Opportunities for professional growth, responsibility, and career development as the company scales.
- Equipment, meals, transportation, and other office benefits covered by the company.
