About this role
Chryselys seeks a Generative AI Agent Engineer in Hyderabad, India, to design, build and optimize LLM-powered agents that reason, use tools and reliably complete complex workflows. The full-time role focuses on agent behavior, tool design, orchestration, evaluation and performance.
Responsibilities
- Design agent architectures, workflows, instructions, decision logic and collaboration patterns. Build tools with clear schemas, inputs, outputs, permissions, error handling and user feedback.
- Develop agents that plan, retrieve information, call APIs and complete multi-step tasks. Design memory, context, retrieval and summarization strategies; build multi-agent interactions where appropriate.
- Improve accuracy, task completion, latency, cost, consistency and user experience. Create evaluations, test scenarios, benchmarks and quality metrics, and analyze traces and failures involving prompts, tools, routing, context or orchestration.
- Implement guardrails for safety, security, hallucination reduction, authorization and human-in-the-loop escalation. Work with application and platform engineers on production integration and establish reusable development practices.
Requirements
At least 3 years of experience in software engineering, machine learning or GenAI application development, plus hands-on experience designing and deploying LLM-powered agents or agentic workflows. Candidates should understand prompting, tool calling, structured outputs, planning, RAG, memory and context management; have experience designing reliable APIs and agent tools; and be proficient in Python, TypeScript or a comparable language. Experience with an agent framework such as LangGraph or LangChain, familiarity with AWS, and experience in evaluation, observability, tracing, experimentation and performance optimization are required. The role also calls for understanding of distributed systems, APIs, asynchronous workflows and production software practices. AWS AgentCore is preferred.
Preferred
Experience with multi-agent systems, A2A or MCP protocols, and enterprise or regulated-industry workflows; familiarity with model routing, prompt versioning, fine-tuning, guardrails and human-in-the-loop patterns; understanding of identity, authorization, data privacy and secure tool execution; and the ability to turn business processes into reliable, measurable agent workflows.