**METEORO platform **
📍Remote
Responsibilities:
- Generative Model Development: Design, train, fine-tune, evaluate, and deploy generative models for applications involving images, video, text, and multimodal data.
- Image and Video Generation: Enhance visual quality, prompt adherence, spatial alignment, character/object consistency, and temporal coherence in generative video.
- LLM Control Research: Investigate and implement methods for controlling and constraining LLM behavior, including instruction tuning, structured generative creation, constrained decoding, tool use, safety mechanisms, preference optimization, and steered generation.
- LLM Evaluation: Develop evaluation systems for instruction following, factuality, safety, output consistency, tool-use accuracy, and robustness against prompt injection or unintended behaviors.
- Model Optimization: Optimize training and inference pipelines for latency, throughput, memory usage, generation quality, and infrastructure costs.
- Pipeline Construction: Build scalable pipelines for data processing, training, fine-tuning, evaluation, and real-time inference.
- Production Integration: Collaborate with software and data engineers to integrate generative models into existing products and production systems.
- MLOps: Automate model testing, deployment, monitoring, rollback, retraining, and continuous evaluation.
Research and Experimentation: Analyze the latest research in generative AI, reproduce promising methods, and determine their practical value for production environments.
Required Skills:
- Experience: At least 3 years of professional experience in machine learning and software engineering.
- Generative AI: Extensive hands-on experience developing or deploying generative AI models.
- Image and Video Models: Deep understanding of diffusion models and state-of-the-art image or video generation architectures.
- Large Language Models (LLMs): Experience with LLM prompting, fine-tuning, evaluation, structured outputs, tool calling, and behavior control techniques.
- ML Frameworks: Hands-on experience with PyTorch or TensorFlow.
- Production ML: Experience migrating models from prototype to production and operating reliable inference services.
- Infrastructure: Experience deploying models on GPU servers without an operating system and on cloud platforms.
- Engineering: Deep Python and software engineering skills, including testing, monitoring, debugging, and performance optimization.
Nice to have:
- Experience with guided generation, instruction tuning, RLHF, DPO, reinforcement learning, or other preference optimization methods.
- Experience with constrained decoding, function calling, agent systems, safety mechanisms, red teaming, and automated LLM evaluation.
- Experience with multimodal models combining text, images, video, or audio.
- Proven ability to improve temporal coherence, spatial alignment, identity preservation, and motion quality in generated video.
- Experience optimizing GPU computations using quantization, batching, caching, compilation, or distributed execution.
- Experience with MLOps platforms, model registries, experiment tracking, and observability systems.
- Familiarity with CI/CD pipelines and Infrastructure as Code.
- Proven experience operating high-throughput, low-latency production systems.
