ml engineer for generative video

Higgsfield AI is a generative AI company developing AI-powered video creation and next-generation creative tools. Its products serve more than 25 million users worldwide, support millions of daily generations, and power hundreds of Fortune 500 brands.

• Own applied ML projects end to end, from problem definition and data collection to production deployment and monitoring;
• Build multimodal classification and detection systems for images, video, text, and metadata;
• Develop solutions for NSFW detection, intellectual property and character recognition, content policy enforcement, and related trust and safety use cases;
• Fine-tune and evaluate vision-language models, classifiers, embedding models, and other relevant architectures;
• Build training and evaluation datasets using human labeling, synthetic data, hard-negative mining, and active learning;
• Define evaluation frameworks that reflect real production scenarios;
• Design and optimize inference pipelines for high throughput, low latency, reliability, and cost efficiency;
• Establish production monitoring for model quality, data drift, policy coverage, false positives, and false negatives;
• Run experiments and analyze the impact of ML systems on user experience, platform safety, conversion, retention, generation success rate, and operational costs;
• Work with Product, Engineering, Legal, Policy, and Operations teams to translate business and policy requirements into scalable technical systems;
• Make build-versus-buy decisions and combine internal models, third-party solutions, and rule-based systems where appropriate;
• Contribute to the architecture and technical direction of the applied ML platform;
• Own production metrics including precision and recall, false-positive and false-negative rates, appeal and moderation reversal rates, generation success and completion rates, inference latency, system availability, classification or generation cost, manual review volume, coverage, retention, engagement, and conversion;
• Independently take ownership of a high-impact applied ML problem, establish an evaluation baseline, deploy an initial production solution, and create a measurable improvement loop;
• Help build a scalable content intelligence and trust and safety platform supporting new models, products, policies, and markets.

• 5+ Years of experience in machine learning with significant experience deploying ML systems into production;
• Strong experience with computer vision, multimodal machine learning, content understanding, recommendation, ranking, fraud detection, trust and safety, or a related applied ML domain;
• Proven ability to independently own complex ML projects from an ambiguous business problem through production launch;
• Strong understanding of model evaluation, including precision and recall trade-offs, threshold selection, calibration, class imbalance, and cost-sensitive decision-making;
• Experience building datasets, labeling workflows, evaluation sets, and feedback loops for continuous model improvement;
• Experience deploying and operating models at scale, including inference optimization, monitoring, retraining, and incident response;
• Strong Python skills and experience with modern ML frameworks such as PyTorch;
• Ability to work with large-scale data and production systems;
• Strong product judgment and understanding of how model performance connects to user experience and business outcomes;
• Ability to communicate technical trade-offs clearly to technical and non-technical stakeholders;
• Nice to have: experience with trust and safety, content moderation, copyright or intellectual property detection, generative image or video models, vision-language models, embeddings, similarity search, perceptual hashing, retrieval systems, human-in-the-loop review and annotation systems, adversarial behavior, model evasion, abuse patterns, changing content distributions, or fast-moving startup environments.

• Competitive base salary in USD;
• Equity participation in the company’s stock option program;
• Relocation from anywhere is provided.