Senior Software Consultant - Agentic AI - Morning Shift

Company Overview
10Pearls is an award-winning end-to-end digital innovation company that helps businesses imagine and build the future. We are proud to announce that 10Pearls was named as winner of the Best Tech Work Culture Timmy Award in Washington DC by Tech in Motion, recognized on the Inc. 5000 Fastest-Growing Companies List, and was ranked the #1 Most Diverse Midsize Company in Greater Washington. We partner with businesses to help them transform, scale, and accelerate by adopting digital and exponential technologies. Our work has ranged from creating highly usable, secure digital experiences, mobile and software products, to helping businesses modernize through cloud adoption and development and the digitalization of their business processes. Our clientele is highly diverse, including Global 1000 enterprises, mid-market businesses, and high-growth start-ups. But those are just the facts. What makes us unique is that we have true heart and soul. We have a strong focus on a double bottom line and actively support and engage with the communities where we live and work to make the world a better place. In a nutshell, we believe in doing well, while doing good, and know how to balance the two.

Role
10Pearls is seeking an Agentic AI Engineer to design and build intelligent AI-driven systems powered by Large Language Models (LLMs). The ideal candidate will work on developing agent-based AI workflows, integrating LLMs into real-world applications, and building scalable AI-powered solutions. This role involves close collaboration with product, engineering, and data teams to deliver innovative AI capabilities.

Responsibilities
• Agent nodes — implement LangGraph StateGraph nodes (plan, retrieve, tool-call, simulate, propose, respond) with clear state transitions and failure handling
• Tool integration — wire agents to L3 Tool Catalog tools and model-as-tool endpoints; implement function-calling contracts and structured-output parsing
• Prompt engineering — author system prompts, few-shot examples, and persona configurations; maintain prompt versions in source control
• Guardrails — implement input and output rails (injection detection, PII redaction, topic steering, refusal behaviours) in line with the Agentic AI Lead’s design
• Eval authoring — write deterministic eval suites, regression tests, and tool-selection accuracy tests; measure and reduce variance
• LLM serving integration — integrate with vLLM (or a mock provider in dev); handle timeouts, retries, and graceful degradation
• Streaming UX — implement SSE streaming of thinking, tool calls, and responses to the L0 frontend
• Observability — instrument agent runs with OpenTelemetry; surface tool-call traces, token usage, latency, and eval metrics

Required Qualifications
• 3–5 years in backend or ML engineering with at least 12 months building LLM-powered features in production
• Hands-on experience with LangGraph, LangChain, LlamaIndex, or comparable agent frameworks
• Strong Python (FastAPI, AsyncIO, Pydantic) and working prompt engineering skills
• Working knowledge of LLM inference serving — vLLM, TGI, or OpenAI-compatible APIs
• Experience with function-calling and structured-output patterns across modern LLMs
• Exposure to eval methodology — regression tests, gold sets, LLM-as-judge
• Solid software-engineering discipline — testing, code review, CI hygiene

Nice to Have
• Experience with NeMo Guardrails, Guardrails.ai, or equivalent safety frameworks
• RAG pipelines with vector stores (pgvector, Qdrant, Weaviate)
• MCP (Model Context Protocol) or similar tool-discovery protocols
• Fine-tuning or LoRA adaptation of open-weights models
• Knowledge-graph or ontology-driven agent design