Research Scientist - Audio [33341]

Stealth Startup; San Francisco (CA); On-site; USD 180,000 - 280,000; Full time

Job summary

Stealth Startup is seeking a Founding Machine Learning Research Engineer to advance real-time AI by exploring state-of-the-art approaches across LLMs, speech models, and multimodal AI for human-like voice agents in complex environments.

You will design evaluation methodologies, prototype innovative systems, and improve reasoning, latency, and conversational intelligence. Your research will influence production systems and drive scalable deployment for thousands of customers.

Qualifications

• Strong machine learning research background in areas such as LLM pre-training and post-training.
• Speech recognition (ASR).
• Deep understanding of modern machine learning architectures and algorithms.
• Strong proficiency with PyTorch.

Responsibilities

• Research and develop novel techniques across LLMs, speech models, and multimodal AI.
• Improve reasoning capabilities, latency, conversational intelligence, and system robustness.
• Explore new training paradigms, architectures, and inference techniques.
• Design, train, and iterate on machine learning models and supporting pipelines.
• Rapidly prototype research ideas and validate them through experimentation.
• Optimize models for production performance and scalability.
• Design innovative evaluation frameworks for conversational AI.
• Create datasets, benchmarks, and metrics that reflect real-world customer interactions.
• Measure model quality across latency, reasoning, speech quality, and user satisfaction.
• Partner closely with engineering teams to deploy research innovations into production.
• Optimize models for reliability, scalability, and operational efficiency.
• Develop methodologies that incorporate human evaluation into model training.
• Improve conversational quality through reinforcement learning and human feedback.
• Build feedback loops that continuously enhance AI performance.
• Stay current with cutting-edge ML research.
• Evaluate emerging techniques and incorporate them into production systems.
• Contribute to the long-term technical direction of AI infrastructure.

Skills

LLM research

Speech recognition

Mathematical foundations

Modern ML architectures

Reinforcement learning

Education

Master's degree in CS/AI/ML or related field

PhD preferred or equivalent research experience

Tools

PyTorch

Job description

This is a research-driven, high-impact opportunity for machine learning researchers passionate about advancing real-time AI.

As a Founding Machine Learning Research Engineer, you'll explore state-of-the-art approaches across large language models (LLMs), speech models, and multimodal AI to build human-like voice agents capable of operating in complex real-world environments.

You'll design new evaluation methodologies, prototype innovative systems, and improve reasoning, latency, and conversational intelligence. Your research will directly influence production systems, bridging cutting-edge machine learning research with large-scale deployment.

If you enjoy solving open-ended ML challenges, experimenting rapidly, and building AI systems used by thousands of customers, this role offers a unique opportunity to shape the future of conversational AI.

Key Responsibilities

Research & Experimentation

• Research and develop novel techniques across LLMs, speech models, and multimodal AI.
• Improve reasoning capabilities, response latency, conversational quality, and system robustness.
• Explore new training paradigms, architectures, and inference techniques.
• Design, train, and iterate on machine learning models and supporting pipelines.
• Rapidly prototype research ideas and validate them through experimentation.
• Optimize models for production performance and scalability.
• Design innovative evaluation frameworks for conversational AI.
• Create datasets, benchmarks, and metrics that reflect real-world customer interactions.
• Measure model quality across latency, reasoning, speech quality, and user satisfaction.

Production Impact

• Partner closely with engineering teams to deploy research innovations into production.
• Optimize models for reliability, scalability, and operational efficiency.

Human Feedback Systems

• Develop methodologies that incorporate human evaluation into model training.
• Improve conversational quality through reinforcement learning and human feedback.
• Build feedback loops that continuously enhance AI performance.
• Stay current with cutting-edge ML research.
• Evaluate emerging techniques and incorporate them into production systems.
• Contribute to the long-term technical direction of AI infrastructure.

Minimum Qualifications

• Strong machine learning research background in areas such as:
• LLM pre-training and post-training
• Speech recognition (ASR)
• Deep understanding of modern machine learning architectures and algorithms.
• Strong proficiency with PyTorch.
• Excellent mathematical foundation in machine learning and deep learning.
• Master's degree in Computer Science, Artificial Intelligence, Machine Learning, or a related field required.
• PhD preferred, or equivalent research-level industry experience.