Staff ML - GenAI Engineer, Voice & Speech

What matters most

• Architecting Scalable Voice & Audio ML Infrastructure

  The role requires designing robust, production-grade platforms for voice and audio GenAI. You must demonstrate the ability to build resilient services that handle large-scale data and traffic, ensuring that complex ML models are accessible and performant for cross-functional teams.

• Bridging Research and Practical B2B Application

  You will act as a technical leader who translates sophisticated ML research into tangible, customer-facing features. This involves consulting with product teams on ML patterns, managing the experimental nature of AI, and ensuring that internal tools are intuitive and reliable for non-ML specialists.

• Leadership and Cross-Functional Mentorship

  As a Staff Engineer, you are expected to coach and collaborate across the organization. You must be able to guide teams through the lifecycle of AI-powered features, from initial design to deployment, while fostering a culture of technical excellence and best practices.

Likely challenges to think through

• Balancing Low-Latency Requirements with Model Complexity

  Implementing real-time audio models (ASR, speech synthesis) requires strict latency management. Prepare to discuss how you optimize inference pipelines and manage trade-offs between model accuracy and real-time performance in distributed cloud environments.

• Managing Sensitive Data in AI Workflows

  The role involves handling large-scale data, potentially including PHI/PII. You will face the challenge of building AI systems that remain compliant and secure while still allowing for effective model training and evaluation. Prepare to discuss your approach to data privacy and secure ML infrastructure.

• Navigating Distributed Systems and Multi-Tenant Complexity

  Deploying ML-driven applications in a multi-tenant B2B environment introduces significant complexity in resource management and isolation. Be ready to explain how you design for scalability and observability using tools like Kubernetes, Dagster, or MLFlow to ensure consistent performance across different tenants.

Topics to prepare

• Distributed Systems for ML at Scale

  The role requires managing TBs of data and multi-tenant production environments. Exercise: Review architectural patterns for building resilient, observable ML services that handle high-concurrency requests and large-scale data processing in cloud environments.

• Voice and Audio GenAI Modalities

  Focus on real-time ASR, speech synthesis, and audio alignment. Exercise: Research current state-of-the-art architectures for low-latency audio pipelines and interruption detection to prepare for discussions on optimizing voice-based user experiences.

• ML Platform Engineering and Lifecycle

  The position involves building internal platforms to democratize AI. Exercise: Design a conceptual framework for an internal ML platform that supports model versioning, automated evaluation, and deployment, specifically addressing the needs of cross-functional teams.

Skills to revise

• LLM Evaluation and Fine-tuning

  Mandatory expertise in evaluating LLM performance and fine-tuning models. Exercise: Define a robust evaluation strategy for a RAG-based application, including metrics for accuracy, latency, and hallucination detection.

• Workflow Orchestration (Dagster/KubeFlow)

  Experience with workflow engines is required for productionizing ML. Exercise: Create a sample pipeline configuration using Dagster or KubeFlow that automates data ingestion, model training, and deployment steps for an audio processing task.

• Kubernetes and Container Orchestration

  Preferred experience with GKE and container patterns. Exercise: Practice defining Kubernetes manifests or Helm charts for deploying a Triton Inference Server, focusing on resource management and scaling strategies for high-traffic audio models.

• Prepare a technical case study on high-scale audio/voice AI

  Draft a detailed walkthrough of a production-grade system you designed that involved real-time audio processing, such as ASR or speech synthesis. Focus on how you handled low-latency requirements, multi-tenancy, and large-scale data ingestion in a cloud environment.

• Develop a framework for LLM and audio model evaluation

  Outline your methodology for evaluating GenAI models, specifically addressing how you measure quality, drift, and performance in production. Be prepared to discuss how you balance model accuracy with the constraints of sensitive data like PHI or PII.

• Map your experience to Weave's ML stack

  Create a summary of your hands-on experience with the specific tools mentioned in the job description, including Python, Dagster, MLFlow, KubeFlow, and Triton Server. Highlight instances where you used these to build internal platforms or SDKs for other engineering teams.

• Formulate questions on cross-functional AI leadership

  Prepare thoughtful questions for the interviewers regarding how the ML team currently consults with product and backend teams on AI tradeoffs. Ask about the biggest challenges they face in democratizing AI access for developers without data science backgrounds.

• Review distributed systems and cloud architecture patterns

  Refresh your knowledge on designing resilient, observable services in AWS or GCP. Prepare to discuss your experience with Kubernetes/GKE and how you have implemented container-based workflows to support high-traffic, data-intensive applications.