Senior Full Stack AI Engineer — Agentic at JSM Consulting Inc. in Pune, Maharashtra

Job description

• Location and working model: Pune, India — hybrid. Minimum three hours of daily overlap with US Central time.
• Experience: 5+ years in engineering, including at least 1 year building LLM-backed products that ran in production.
• Role overview
• We is building enterprise-grade software solutions that combine modern web engineering, cloud platforms and production-grade AI capabilities. This is a senior hands-on role anchoring agentic development: you will design and build LLM-powered agents, establish the reusable engineering patterns other engineers build against, and set the standard for how non-deterministic systems are evaluated. Mentoring and design contribution beyond your own features are core to the role, not occasional extras.

Responsibilities

• Anchor the hands-on build of LLM-powered agents — reasoning loops, planning, tool and function calling, retries, fallbacks and guardrails.
• Establish reusable agent engineering patterns and shared libraries that other engineers build against.
• Define and implement evaluation for agent behaviour: how quality is proven before release and how regression is detected afterwards.
• Build retrieval and structured data access over large corpora to support agent reasoning. ● Instrument agents for observability — tracing, token accounting and cost attribution.
• Contribute to design reviews across the wider engineering group and review other engineers' agent work.
• Mentor engineers on agentic patterns and on responsible AI-assisted development, defining review, testing and traceability standards for AI-generated changes so unverified or architecture-breaking code does not reach the codebase.
• Integrate agent capability with surrounding platform services, including administration, access control and workflow.

Requirements

• Deep production Python — async patterns, queues, structured outputs, and the discipline to make non-deterministic systems testable.
• Hands-on experience with an agent framework (for example LangGraph, LlamaIndex, CrewAI or AutoGen, or an in-house equivalent) and with function and tool calling.
• Prompt design paired with evaluation: you can explain how you measured whether an agent worked, not only how you built it.
• Retrieval-augmented generation or structured retrieval over a large corpus.
• Observability for LLM systems — tracing, token accounting and regression suites for agent behaviour.
• Demonstrated mentoring: pairing, design review and visibly raising the standard of the engineers around you.
• Disciplined use of AI coding assistants within defined engineering and security guardrails: you independently validate generated code, verify it through unit, integration and regression tests, check maintainability, architecture alignment, security, privacy, licensing and performance, and never treat generated output as production-ready without human review. You can explain and defend any code you submit.
• Comfortable with ambiguity — agent behaviour is established empirically rather than specified up front.
• Full stack range — able to build the interfaces that expose agent capability.
• QA or test-automation domain knowledge.
• Experience with browser automation or DOM analysis libraries. ● Cost and performance optimization for LLM workloads.

Skills

Required

• Deep production Python
• Async patterns
• Queues
• Structured outputs
• Function and tool calling
• Prompt design
• Evaluation of agent behaviour
• Retrieval-augmented generation
• Structured retrieval
• Observability for LLM systems
• Tracing
• Token accounting

Preferred

• Full stack development
• QA or test-automation
• Browser automation
• DOM analysis libraries