Software Engineer, AI - Python

Telnyx builds private global connectivity networks and API-driven communication infrastructure for people, devices, and applications. The company is seeking a Software Engineer to design, build, and operate AI-native backend systems and multi-agent architectures that support go-to-market workflows. The role includes developing Python services, integrating external business systems, operating Kubernetes deployments, and monitoring production AI systems.

Responsibilities:

• Design and build multi-agent AI systems in Python that handle complex, multi-step business workflows - qualification, email generation, routing, enrichment, and outbound orchestration
• Architect model-agnostic abstraction layers that decouple business logic from LLM providers, enabling flexibility across Claude, GPT, and open-source models
• Build and operate backend services (FastAPI/Flask) deployed on Kubernetes with CI/CD, managing the full lifecycle from deployment configuration to production reliability
• Design tool-use patterns for AI agents - structured function calling, multi-step reasoning, state management across conversation turns, and graceful handling of model failures
• Build integrations across external systems (CRM, enrichment APIs, outreach platforms, Slack) with proper error handling, retries, rate limiting, and data contracts
• Instrument and monitor AI systems in production — build observability into agent behavior, track success rates, detect regressions, and debug non-deterministic failures
• Design and run experiments (A/B tests, prompt variations, model comparisons) with proper evaluation infrastructure to measure what's actually working

Requirements:

• 2+ years of software engineering experience building backend services in Python
• Production experience building multi-step AI agent systems — stateful workflows where models make decisions, call tools, and operate across multiple turns, not single-shot API wrappers
• Strong understanding of LLM internals as they affect system design: context window management, token budgets, cost/latency/capability tradeoffs across models, structured outputs, and strategies for handling hallucination and refusals
• Experience testing and evaluating non-deterministic AI systems — you understand that assert output == expected doesn't work and have built or used alternatives
• Solid software architecture fundamentals: API design, state management, fault tolerance, and graceful degradation when upstream services fail
• Production experience with containerized deployments (Docker, Kubernetes) and CI/CD pipelines
• Experience integrating with external APIs at scale — auth flows, rate limiting, retries, data normalization, and managing the operational complexity of multiple third-party dependencies
• Proficiency with SQL and data systems for building targeting, enrichment, and analytics pipelines
• Built observability into production systems — structured logging, tracing, alerting, and monitoring that you actually use to debug issues
• High ownership: you deploy your own code, investigate your own incidents, and close the loop between what you shipped and how it performs
• Experience with specific GTM/RevOps systems (Salesforce, Apollo, Lusha, enrichment providers) or similar complex business platforms
• Background in growth engineering, marketing automation, or revenue operations tooling
• Experience with Slack bot development or conversational AI interfaces
• Contributions to or experience with open-source AI agent frameworks
• Familiarity with ArgoCD, StatefulSets, or Kubernetes operations beyond basic deployments