Senior CX Engineer

Caylent is an AI-first cloud services company that helps organizations turn ambitious ideas into meaningful business impact. As an AWS Premier Tier Services Partner and a charter member of Anthropic’s Claude Partner Network, we combine deep expertise in AWS, artificial intelligence, and Anthropic’s Claude platform to help customers modernize their technology, build intelligent products, and move AI from experimentation into production.

Our capabilities span generative and agentic AI, cloud migration and modernization, cloud-native application development, data and analytics, Dev Ops, managed services, security and compliance, and customer experience transformation. At Caylent, our people always come first.

We are a fully remote global company with employees in Canada, the United States and Latin America. We celebrate the culture of each of our team members and foster a community of technological curiosity. Come talk to us to learn more about what it means to be a Caylien!

As a Senior CXE Engineer, you are a hands‑on builder on enterprise contact center transformations. You will spend the majority of your time in code and in the AWS console: building contact flows and AI agents, writing Lambda functions and CDK stacks, developing React-based agent and admin tooling, and debugging production issues on systems that serve tens of thousands of agents and millions of customer interactions.

You work within architectures set by practice leads, and you’re expected to pressure‑test them, propose better patterns, and own the implementation quality of your workstream from first commit through production cutover. This role blends deep Amazon Connect platform engineering, generative AI development on Bedrock and Claude, and full‑stack serverless work, with regular direct interaction with customer engineering teams.

• Build AI‑native self‑service using Amazon Connect AI agents, Amazon Q in Connect, and Amazon Bedrock (Claude model family), implementing multi‑agent orchestration patterns, orchestrator agents delegating to task‑specific agents with escalation paths to human queues
• Engineer prompts and tool‑calling integrations at production scale:
  Search Profiles lookups, case creation, external API tools, authentication flows, and guardrails against hallucination and verification loops
• Implement barge‑in handling, fallback, and error‑recovery behavior for voice AI, and validate it with turn‑by‑turn trace analysis
• Optimize inference cost and latency hands‑on: model tier selection (Haiku vs. Sonnet vs. Opus), prompt caching, and token budgeting against containment‑ rate targets
• Diagnose AI agent performance using observability tooling (agent spans, and Cloud Watch) that correlates contact flow logs, conversation transcripts, AI agent spans, tool executions, and token usage to isolate latency, cost, and quality issues
• Build conversational experiences where intent‑based NLU is the right fit, including multilingual support and SSML‑tuned voice interactions

• Build contact flows that integrate AI agents, Lambda‑backed business logic, Customer Profiles, and dynamic data‑driven routing
• Implement Amazon Connect Cases and Tasks solutions: case templates, queue‑based work allocation, SLA/TAT monitoring, approval workflows, and migrations from legacy case systems (e.g., consolidating hundreds of service categories and thousands of business rules into a unified Cases model)
• Configure and operationalize Contact Lens for real‑time sentiment, post‑contact summaries, category rules, and quality management
• Implement proficiency‑based and attribute‑driven routing, outbound campaigns (including answering machine detection and compliance controls), and workforce management (WFM) integrations for forecasting, scheduling, and adherence (FCS)
• Build out the Amazon Connect analytics data lake (Glue, Lake Formation, Athena) and Quick Sight dashboards for real‑time operations and executive reporting
• Support legacy platform migrations (Avaya, Genesys, Cisco, Oracle, Salesforce): flow conversion, data migration, parallel‑run validation, and production…