We're building the infrastructure that lets people talk to Claude,real-time, bidirectional voice conversations that feel natural, responsive, and safe. This is foundational work for how millions of people will interact with AI.

The Voice Platform team designs and operates the serving systems, streaming pipelines, and APIs that bring Anthropic's audio models from research into production across Claude.ai, our mobile apps, and the Anthropic API. You'll work at the intersection of real-time media, low-latency inference, and distributed systems,building infrastructure where every millisecond of latency is felt by the user.

We partner closely with the Audio research team, who train the speech understanding and generation models, and with product teams shipping voice experiences to users. Your job is to make those models fast, reliable, and delightful to talk to at scale.

Responsibilities

• Design and build the real-time streaming infrastructure that powers voice conversations with Claude,ingesting microphone audio, orchestrating model inference, and streaming synthesized speech back with minimal latency

• Build low-latency serving systems for speech models, optimizing time-to-first-audio and end-to-end conversational responsiveness

• Develop the public and internal APIs that expose voice capabilities to Claude.ai, mobile clients, and third-party developers

• Own the audio transport layer,codecs, jitter buffers, adaptive bitrate, packet loss recovery,so conversations stay smooth across unreliable networks

• Build observability and quality-measurement systems for voice: latency distributions, audio quality metrics, interruption handling, and turn-taking accuracy

• Partner with Audio research to move new model architectures from experiment to production, and feed real-world performance data back into research

• Collaborate with mobile and product engineering on client-side audio capture, playback, and the end-to-end user experience

You may be a good fit if you

• Have 6+ years of experience building distributed systems, real-time infrastructure, or platform services at scale

• Have shipped production systems where latency is measured in tens of milliseconds and users notice when you miss

• Are comfortable working across the stack,from transport protocols and serving infrastructure up to the APIs product teams build on

• Are results-oriented, with a bias toward flexibility and impact

• Pick up slack, even if it goes outside your job description

• Enjoy pair programming (we love to pair!)

• Care about the societal impacts of voice AI and want to help shape how these systems are developed responsibly

• Are comfortable with ambiguity,voice is a fast-moving space, and you'll help define the architecture as we learn what works

Strong candidates may also have experience with

• Real-time media protocols and stacks: WebRTC, RTP, gRPC bidirectional streaming, or WebSockets at scale

• Audio engineering fundamentals: codecs (Opus, AAC), voice activity detection, echo cancellation, jitter buffering, or audio DSP

• Low-latency ML inference serving, streaming model outputs, or GPU-based serving infrastructure

• Telephony, live streaming, video conferencing, or voice assistant platforms

• Mobile audio pipelines on iOS (AVAudioEngine, AudioUnits) or Android (Oboe, AAudio)

• Working alongside ML researchers to productionize models,speech experience is a plus but not required

Representative projects

• Driving time-to-first-audio below human perceptual thresholds by co-designing the serving pipeline with the Audio research team

• Building a streaming inference orchestrator that interleaves speech recognition, LLM reasoning, and speech synthesis with overlapping execution

• Designing the voice mode API surface for the Anthropic API so developers can build their own voice agents on Claude

• Implementing graceful barge-in and interruption handling so users can cut Claude off mid-sentence naturally

• Instrumenting end-to-end audio quality metrics and building dashboards that catch regressions before users do

XML job scraping automation by YubHub

Your work will touch the full GenAI inference stack , from kernels and runtimes to orchestration and memory management. You will contribute to the design and implementation of the inference engine, and collaborate on model-serving stack optimized for large-scale LLMs inference.

Key responsibilities include:

• Collaborating with researchers to bring new model architectures or features (sparsity, activation compression, mixture-of-experts) into the engine
• Optimizing for latency, throughput, memory efficiency, and hardware utilization across GPUs, and accelerators
• Building and maintaining instrumentation, profiling, and tracing tooling to uncover bottlenecks and guide optimizations
• Developing and enhancing scalable routing, batching, scheduling, memory management, and dynamic loading mechanisms for inference workloads
• Supporting reliability, reproducibility, and fault tolerance in the inference pipelines, including A/B launches, rollback, and model versioning
• Integrating with federated, distributed inference infrastructure – orchestrate across nodes, balance load, handle communication overhead
• Collaborating cross-functionally: with platform engineers, cloud infrastructure, and security/compliance teams
• Documenting and sharing learnings, contributing to internal best practices and open-source efforts when possible

Requirements include:

• BS/MS/PhD in Computer Science, or a related field
• Strong software engineering background (3+ years or equivalent) in performance-critical systems
• Solid understanding of ML inference internals: attention, MLPs, recurrent modules, quantization, sparse operations, etc.
• Hands-on experience with CUDA, GPU programming, and key libraries (cuBLAS, cuDNN, NCCL, etc.)
• Comfortable designing and operating distributed systems, including RPC frameworks, queuing, RPC batching, sharding, memory partitioning
• Demonstrated ability to uncover and solve performance bottlenecks across layers (kernel, memory, networking, scheduler)
• Experience building instrumentation, tracing, and profiling tools for ML models
• Ability to work closely with ML researchers, translate novel model ideas into production systems
• Ownership mindset and eagerness to dive deep into complex system challenges
• Bonus: published research or open-source contributions in ML systems, inference optimization, or model serving

XML job scraping automation by YubHub

You will have the opportunity to contribute directly to research projects yourself and have a direct impact on our ability to scale reward development across domains. You will work closely with researchers and translate ambiguous requirements into well-scoped engineering projects.

To be successful in this role, you should have prior research experience and be excited to work closely with researchers. You should have strong Python skills and experience with ML workflows and data pipelines, and building related infrastructure/tooling/platforms. You should be comfortable working across the stack, ranging from data pipelines to experiment tracking to user-facing tooling.

Strong candidates may also have experience with ML research, building internal tooling and platforms for ML researchers, data quality assessment and pipeline optimization, experiment tracking, evaluation frameworks, or MLOps tooling. They may also have experience with large-scale data processing, Kubernetes, distributed systems, or cloud infrastructure, and familiarity with reinforcement learning or fine-tuning workflows.

XML job scraping automation by YubHub