Data Scientist

• Python Programming (Advanced):. You must be proficient in Python, utilizing object-oriented programming, modern type hinting, and asynchronous patterns.
• Environment & Dependency Management: Familiarity with modern Python package managers like uv, virtual environments (.venv), and pip.
• Command Line Interface (CLI) Proficiency: Ability to navigate CLI tools, as well as general shell scripting
• We are looking for Google CX Developers with at least 2 - 4 years of relevant experience.
• Bachelor's/Master's in Engineering 2-5 years

Generative Agent Design: Shifting from legacy intent-based state machines to generative, goal-oriented architectures (understanding Apps, Agents, Sub-agents, and Sessions within CX Agent Studio).

Prompt Engineering & Context Management: Writing robust system instructions, managing conversational memory, and optimizing LLM context windows for voice interactions.

Dialogflow CX Fundamentals: Understanding the underlying mechanics of Dialogflow CX.

Cloud Compute & Serverless: Deploying agent components, webhook integrations, or backend APIs using Google Cloud Functions or Cloud Run.

Gcloud CLI Mastery: Utilizing gcloud for project configuration and authenticating environments via application-default credentials.

Tool Calling / Function Calling: Designing and registering external APIs ("Tools") that the LLM can invoke to retrieve data, execute backend tasks, or interact with external services.

Data Handling & Payload Parsing: Using utility functions to handle pagination, flatten API responses, and convert complex Protocol Buffers (Protos) into usable data.

Automated Agent Evaluation (Evals): Creating and orchestrating "Golden tests" and automated simulation runs using SCRAPI’s evals module.

Performance Metrics Tracking: Extracting, analyzing, and optimizing agent performance metrics (like real-time latency), which is highly critical for voice voice interactions.

Agentic IDE workflows: Using LLM-assisted development tools (like Gemini CLI or Claude Code) as integrated into the SCRAPI workflow to speed up agent scaffolding and debugging.