Full Stack Engineer (AI focus)

O projekcie: Job Title: Full Stack Engineer (AI focus) The Project:  You will be a key driver in a mission-critical project for our US customer. Your mission is to take an existing Proof of Concept (POC) for ambient listening—the AgentVoice codebase—and refactor it into a robust series of microservices. These services will: - Passively transcribe doctor-patient encounters. - Transform raw transcriptions into highly structured clinical notes. - Generate actionable insights and automated workflow triggers to reduce administrative burden on physicians. This position is initially planned for 6 months starting in May. Further renewal is tied to commercial adoption of what we build. Wymagania: - Python & Backend Ecosystem Core: Senior-level proficiency in Python 3.11+, focusing on asynchronous programming (asyncio) and high-performance API development. Frameworks: Extensive experience with FastAPI or Flask for building scalable microservices. AI Orchestration: Deep knowledge of LangChain or LlamaIndex for managing LLM chains, memory, and retrieval. Data & Processing: Experience with Pydantic for rigorous data validation and SQLAlchemy/Tortoise-ORM for database interactions. Messaging: Familiarity with Kafka and Redis/Valkey for handling asynchronous transcription tasks and background jobs. - React & Frontend Excellence Core: Advanced React 18+ (Hooks, Context API, Suspense) and TypeScript. NextJS/React Router Framework. State Management: Experience with Zustand, TanStack Query (React Query), or Redux Toolkit to manage complex, real-time data streams. UI/UX: Proficiency in Tailwind CSS and component libraries (e.g., Shadcn/UI) to build clean, efficient clinical dashboards. Real-time Data: Experience with WebSockets or Server-Sent Events (SSE) for displaying live transcription updates. - AI Tooling & The "Agentic" Stack LLM Expertise: Hands-on experience with Claude 3.5/4 (Anthropic) and OpenAI models. Model Context Protocol (MCP): Ability to implement and extend MCP servers to connect LLMs with secure medical databases and external tools. Skills & Plugins: Experience building Custom Skills (modular capabilities for AI agents) and Plugins that allow LLMs to execute deterministic code/actions. Prompt Engineering: Advanced skills in system prompting, few-shot learning, and output structuring (JSON mode/Function calling). - Infrastructure & DevOps Cloud: Comfort working in multi-cloud environments (AWS or GCP). Containerization: Mastery of Docker and Kubernetes (K8s) for deploying and scaling microservices. [nice-to-have] CI/CD: Experience with GitLab CI for automated testing and deployment. [nice-to-have] Monitoring: Knowledge of Prometheus, or Grafana for observing AI system performance. - Working Hours & Collaboration US Alignment: Due to our partnership with US companies, this role requires significant daily overlap with the US Eastern Time zone (typically a -6 hour difference from Poland). Shift Schedule: Candidates must be comfortable beign on later hours (e.g., 2:00 PM – 7:00 PM or 4:00 PM – 8:00 PM CET) to facilitate real-time collaboration, stand-ups, and rapid feedback loops with the US engineering team. Flexibility: While we prioritize a healthy work-life balance and provide flexibility, the ability to be available during late afternoon and evening hours in Poland is a core requirement for this position to ensure project synchronization and commercial success. Codzienne zadania: - Service Evolution: Architect and refactor the AgentVoice POC into a scalable, resilient microservices architecture. - Agentic Frameworks: Design and implement Custom Skills and Plugins that enable LLMs to execute deterministic code and medical actions. - MCP Integration: Build and extend Model Context Protocol (MCP) servers to securely bridge LLMs with clinical databases and external healthcare tools. - Prompt Engineering: Develop and fine-tune complex system prompts using few-shot learning and JSON-mode output to ensure rigorous medical accuracy. - Backend: Build high-performance, asynchronous APIs using Python 3.11+ and FastAPI/Flask to process real-time medical data. - Frontend: Create responsive clinical dashboards using React 18 and TypeScript, allowing doctors to review and edit AI-generated insights effortlessly. - Real-time Systems: Implement WebSockets or SSE for live, low-latency transcription streaming during patient encounters. - Data Integrity: Use Pydantic for strict validation and manage database interactions via SQLAlchemy or Tortoise-ORM. - Transcription Orchestration: Manage pipelines converting passive audio into text and then into structured clinical notes (diagnoses, plans) via LangChain/LlamaIndex. - Cloud & DevOps: Deploy and scale services across GCP and AWS using Docker and Kubernetes (K8s). - Reliability: Monitor AI performance and "human-centric" reliability using Prometheus, or Grafana. - Rapid Prototyping: Continuously validate and integrate "bleeding-edge" AI libraries to keep the product at the forefront of the industry.