Research Scientist, Gen AI & User Representation Learning
Predactiv · San Francisco Bay Area · 2 mo ago
RemoteRemoteEngineeringFull-time
Responsibilities
- Conduct Applied AI Research
- Research and develop novel machine learning algorithms for user representation learning, semantic embeddings, and foundation-model applications.
- Design, prototype, evaluate, and deploy transformer-based generative AI solutions from research through deployment.
- Develop scalable representation learning techniques using transformers, contrastive learning, self-supervised learning, and retrieval-based architectures.
- Investigate multimodal learning approaches that jointly model structured, behavioral, textual, and other heterogeneous data.
- Build Large-Scale AI Systems
- Train and evaluate models using large-scale behavioral, transactional, social, temporal, and content datasets.
- Build embedding models, retrieval systems, vector databases, and semantic search pipelines.
- Collaborate with platform and infrastructure engineers to deploy production-quality AI models.
- Design rigorous offline and online evaluation methodologies and establish reproducible benchmarking pipelines.
- Collaborate Across Teams
- Work closely with product, engineering, and domain experts to identify impactful research opportunities.
- Translate ambiguous business problems into measurable machine learning objectives.
- Communicate research findings clearly to both technical and non-technical audiences.
- Contribute to the long-term AI research roadmap and technical strategy.
Requirements
- Education
- PhD (completed or near completion) in Computer Science, Machine Learning, Artificial Intelligence, Statistics, or a related quantitative discipline.
- Equivalent industrial research experience will also be considered.
- Technical Expertise
- Strong background in one or more of the following:
- Deep Learning
- Representation Learning
- Transformer architectures
- Generative AI Models
- Contrastive Learning
- Self-supervised Learning
- Embedding Models
- Retrieval-Augmented Generation (RAG)
- Vector Search
- Semantic Search
- Information Retrieval
- Experience with:
- Python
- PyTorch (preferred) or JAX
- Large-scale distributed data processing
- Model experimentation and evaluation
- End-to-end machine learning system development
- GPU Computing
- NVIDIA GPU architecture and CUDA programming fundamentals
- Mixed precision training (FP16/BF16/FP8)
- Profiling and optimizing GPU utilization, communication overhead, and training throughput
- Research Mindset
- Candidates should demonstrate:
- Strong scientific rigor
- Ability to establish meaningful baselines before pursuing more complex models
- Well-designed experiments and reproducible evaluations
- Data-driven decision making
- Intellectual curiosity and independent problem solving