Remote/Full-time, Salary: $4,000 - $8,000+ (depending on experience) About Yolk: Yolk is a VC-backed startup on a bold mission: to build the best sales departments in the world with elite AI coaching available to every team. About the Role: Your code will ingest data from live sales calls, analyze them to find skill gaps, and autonomously trigger personalized AI roleplay sessions to fix those gaps. **What You'll Build (Key Engineering Initiatives) ** Core Simulation Architecture Design the backend systems that power our interactive training loops. You will move us beyond simple request/response cycles into stateful, persistent WebSocket connections that can handle interruptions, context switching, and dynamic feedback in real-time. The Async Evaluation Engine You will architect the analysis pipeline that processes both Real Calls and Simulated Roleplays. This system must scale to handle long-form audio processing, utilizing LLMs to extract structured rubric data (e.g., "Did they ask for budget?") and feeding it into our analytics layer. Observability & Resilience Enhance our "blind spots" in production. You will implement distributed tracing to visualize the full lifecycle of an AI conversation—from packet arrival to LLM inference to audio generation—ensuring we can debug race conditions and latency spikes effectively. The "Gap-to-Game" Orchestrator You will write the algorithms that: ingest performance data from real calls, identify specific skill gaps and assign the correct AI Roleplay scenario to the user. Idempotency & Data Integrity: Ensuring that if something fails, we never corrupt the user's performance metrics. **Technical Challenges You'll Solve ** ● State Management in Stateless Environments: Maintain the precise context of a 20-minute AI conversation when services scale up/down or when connections drop ● Complex LLM Orchestration: Work along our AI research on "Chain-of-thought" pipelines that can branch dynamically based on user input without adding perceptible delay. ● Strict Latency Budgets: Optimizing our Python async loop to handle audio stream chunks and vendor API calls within strict millisecond windows. ● Concurrency at Scale: Designing the system to handle hundreds of concurrent audio streams and LLM contexts without blocking the event loop. ● Core: Python 3.12+ | FastAPI | SQLAlchemy 2.0 (async) | PostgreSQL ● Async & Messaging: RabbitMQ + FastStream | WebSockets | Redis (Optional) ● AI & Data: OpenAI/Anthropic APIs | Vector Stores | Real-time Audio Protocols (WebRTC/RTMP, Pipecat) ● Tooling: uv | ruff | basedpyright (strict) | OpenTelemetry **Must-Have: ** ● 5+ years backend engineering experience, with deep expertise in Python 3.10+. ● Mastery of AsyncIO: You understand the Python event loop inside out. You know how to write non-blocking code that handles concurrent WebSocket connections efficiently. ● API Design Excellence: You build clean, RESTful (or GraphQL/RPC) interfaces that frontend teams love to use. You use Pydantic and strict typing to enforce contracts. ● Database Design: You are comfortable writing complex SQL queries and designing normalized schemas in PostgreSQL/SQLAlchemy. ● Architectural Pattern Knowledge: You understand State Machines, Event-Driven Architecture, and how to decouple services cleanly. (But DevOps handles the heavy lifting): Docker basics, RabbitMQ/FastStream, WebSocket scaling. **🌟 Why You’ll Love Working at Yolk **