Senior Research Engineer
AssemblyAI4 hours ago
Remote, United StatesSenior
Base Salary
$270k - $310k/yr
Responsibilities
- Increase experimental velocity by improving experiment launch, measurement, and debugging workflows.
- Maintain and scale the JAX training framework for large-scale distributed training on TPUs.
- Investigate model-data quality issues, build data-quality tooling, and convert findings into measurable accuracy improvements.
- Analyze production-model accuracy and build evaluation harnesses to identify high-impact improvements.
- Translate research prototypes into production-ready systems and modernize model architectures and infrastructure.
- Optimize production inference for speech language models through serving architecture and techniques including quantization and speculative decoding.
- Investigate and resolve performance bottlenecks from low-level kernels through high-level system design.
- Partner with research, infrastructure, and production engineering teams to identify root causes and ship durable fixes.
Requirements
- Expert-level proficiency with JAX and TPUs, including Flax, Optax, and the XLA compilation pipeline.
- Strong experience optimizing production inference systems, ideally for LLMs or speech models.
- Deep understanding of distributed training at scale, modern deep learning systems, and ML infrastructure best practices.
- Familiarity with continuous batching, KV-cache management, sharding strategies, and quantization.
- Strong Python skills.
- Strong measurement discipline, communication skills, collaboration, and willingness to work across the full machine learning pipeline.
- C++ or Rust experience for kernel-level work is a plus.
- Speech-to-text domain knowledge, including ASR architectures, audio processing, and streaming inference, is a bonus.
Benefits
- The posted U.S. salary range is $270,000-$310,000.
- The company provides a small-team environment with substantial ownership and direct impact.
- The role collaborates across research, infrastructure, and production engineering in a fast-moving organization.
- AssemblyAI states its commitment to equal opportunity and an inclusive workplace.
Categories
About AssemblyAI
AssemblyAI is the best way to build Voice AI apps. We build the industry’s best speech-to-text and speech understanding models, including promptable speech recognition, that serve as critical infrastructure for top Voice AI products like Granola, Dovetail, Ashby, and Cluely. Our speech-to-text models lead the industry in accuracy and quality, so you can build reliable product experiences on top of voice data. And our Speech Understanding models help you go beyond transcription to uncover insights, identify speakers, and highlight key information. We make it simple to get started, with a developer-first API and usage-based pricing that scales effortlessly to millions of hours.