About this role
XenonStack seeks a Machine Learning Engineer, Human-AI Interaction in Mohali, India, to design and improve how AI agents understand, respond and act in enterprise workflows. The role combines prompt engineering, context orchestration and memory architecture to make LLM-based and multi-agent systems accurate, reliable and compliant in production. This is a full-time position with a stated salary of 7 - 10 LPA.
Responsibilities
- Design, test and refine prompts for agent behavior across use cases. Build context pipelines integrating RAG, knowledge graphs, APIs and short-term, long-term and episodic memory.
- Optimize token usage and context allocation in long-running, multi-turn and multi-agent workflows. Create reusable interaction templates and context blueprints for engineering and product teams.
- Implement safety, compliance, tone and brand guardrails. Use execution traces, feedback and automated evaluation to improve responses; A/B test prompt and context variations against accuracy, latency and cost.
- Maintain a library of tested interaction patterns and context-management strategies, and follow developments in multi-agent orchestration and AI interaction design.
Required qualifications
The must-have section specifies 2–4 years in AI/ML engineering, NLP or enterprise software development; the separate job-information field lists work experience as 1–3 years. Candidates need an understanding of LLM architectures, prompt engineering and context-window limits; hands-on experience with RAG pipelines, vector databases and knowledge graph integration; and proficiency in Python and frameworks such as LangChain, LangGraph and LlamaIndex. Familiarity with enterprise AI governance, privacy and compliance standards, and the ability to translate business objectives into structured AI interactions, are also required.
Preferred qualifications
Multi-agent orchestration experience with MCP, A2A messaging or AgentBridge; knowledge of reinforcement learning, including RLHF, RLAIF and reward modeling; exposure to edge AI deployment and quantized inference; and domain knowledge in BFSI, GRC, SOC or FinOps. The company describes opportunities to work on enterprise agentic AI products and with global enterprise customers.