Staff Research Engineer, Ambience Healthcare

• Skilled at interpreting and replicating results from cutting-edge machine learning research
• Experienced in designing experiments, running ablation studies, and ensuring reproducibility
• 4+ years of experience in machine learning research, experimental AI, or applied AI engineering
• Demonstrated ability to replicate, extend, or publish original research
• Hands-on experience fine-tuning large language models and optimizing prompt and embedding strategies
• Proficient with Python and deep learning frameworks such as PyTorch, JAX, and Hugging Face Transformers
• Comfortable with distributed training environments and large-scale model experimentation
• Deep understanding of dataset curation, filtering, and alignment with evaluation goals
• Familiar with human annotation pipelines, ranking models (e.g., RM, RLAIF), and interpretability techniques
• Experienced in building evaluation frameworks tied to real-world task performance
• Thrives in research-driven environments with a commitment to experimentation, documentation, and cross-functional learning
• Excited to prototype, present findings, and build at the frontier of AI advancement
• Able to work alongside clinicians, product managers, and fellow engineers
• Strong communicator who can distill complex ML concepts for diverse audiences
• Passion for healthcare or other mission-driven industries (e.g., education, climate tech)
• Thrives in a fast-paced, early-stage environment; takes extreme ownership of deliverables

Desirable

• As a Staff Research Engineer at Ambience, you will push the boundaries of generative AI by translating cutting-edge research into working prototypes and experimental platforms
• You’ll work closely with fellow researchers, engineers, and product leads to explore novel architectures, fine-tuning methods, evaluation paradigms, and data strategies—helping to define what’s possible with frontier AI models
• Prototype and Advance LLM Systems: Build and benchmark LLM-based systems and agents using open-source and proprietary models. Rapidly prototype new capabilities through fine-tuning, adapters, and reinforcement learning approaches
• Drive Research-First Experimentation: Translate recent academic papers into reproducible experiments, focusing on fine-tuning (e.g., LoRA, QLoRA, DPO), model alignment, and hallucination mitigation techniques. Design clear experiment plans and share findings across the team
• Build and Evolve Evaluation Pipelines: Define evaluation methodologies using human-in-the-loop feedback, synthetic benchmarks, and task-specific metrics. Implement continuous evaluation pipelines to track regressions and breakthroughs
• Shape Data and Training Strategy: Curate datasets via synthetic generation, targeted scraping, and annotation pipelines. Establish practices for discovering failure cases and improving model robustness over time
• Contribute to a Research-Driven Culture: Write research papers, internal memos, and blog posts. Foster a culture of experimentation, documentation, and knowledge-sharing across research and engineering teams