AI Architect
EXL
- Location
- Hybrid (Noida, Uttar Pradesh)
- Employment
- Full-time
- Level
- Senior Level
About the role
EXL is seeking an AI Architect to design and implement generative, goal-oriented agent architectures using Dialogflow CX and Google Cloud Platform. The role focuses on deploying agent components, managing API integrations, and optimizing conversational performance metrics.
Skills
Perks
- Hybrid Work
Full job details
Key Skills:
- Python:
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
- Conversational AI / Agent Architecture:
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,
- Google Cloud Platform(GCP):
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.
- API Integration & Tool Building:
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.
- Testing, Evaluation & CI/CD:
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.
Responsibilities
Key Skills:
- Python:
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
- Conversational AI / Agent Architecture:
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,
- Google Cloud Platform(GCP):
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.
- API Integration & Tool Building:
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.
- Testing, Evaluation & CI/CD:
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.
Qualifications
Key Skills:
- Python:
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
- Conversational AI / Agent Architecture:
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,
- Google Cloud Platform(GCP):
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.
- API Integration & Tool Building:
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.
- Testing, Evaluation & CI/CD:
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.