Conversational AI Engineer: Multi-Agent Code (Remote)

Socket.dev; Austin (TX); On-site; USD 110,000 - 170,000; Full time

Benefits offered by this job

Medical, dental, vision

Tax-advantaged health care accounts

Paid time off

Job summary

TTEC Digital is seeking a Conversational AI Engineer to design and deploy next-generation agentic workflows. You will build modular multi-agent systems and implement custom Python/TypeScript tool callbacks within CX Agent Studio, bridging discovery and front-line operations.

You will also ensure code-first workflows are linted, tested, and synchronized, blending low-code canvas with deterministic logic to prevent hallucinations in production.

Qualifications

• Bachelor’s degree or equivalent practical experience in CS/engineering.
• 3+ years building conversational AI solutions or programmatic AI workflows.
• Strong proficiency in Python or TypeScript/Node.js with RESTful APIs and cloud webhooks.

Responsibilities

• Develop multi-agent architectures with Root/Steering and Sub-agents.
• Write custom code tools and callbacks in Python/TypeScript.
• Leverage CX Agent Studio CLI tools for linting and validation (cxas lint, llm-lint).
• Design structured agent instructions and routing syntax for sub-ag agents.
• Integrate Code Agents with live databases, CRM, and CCaaS via APIs.
• Create automated QA and simulation tests to prevent hallucinations.

Skills

Python

TypeScript/Node.js

RESTful APIs

Git workflows

CX Agent Studio

cxas-scrapi

JSON payloads

Cloud webhooks

BigQuery / Google Cloud Logging

Education

Bachelor’s degree in computer science, engineering field, or equivalent practical experience

Tools

cxas-scrapi workspace/SDK

Python-based agent foundries

Job description

TTEC Digital is seeking a Conversational AI Engineer to design and deploy next-generation agentic workflows. You will build modular multi-agent systems and implement custom Python/TypeScript tool callbacks within CX Agent Studio, bridging discovery and front-line operations.

You will also ensure code-first workflows are linted, tested, and synchronized, blending low-code canvas with deterministic logic to prevent hallucinations in production.