We are currently hiring for the position of Senior AI Engineer for one of our esteemed clients in NY/NJ, Columbus or Dallas (On-Site)
Please find the job details below:
Job title:Senior AI Engineer
Locations: NY/NJ, Columbus or Dallas (On-Site)
Duration: 12+ months contract (extension possible)
The client is a technology organization operating at the scale of a major financial enterprise, building AI agents that serve millions of customers directly. This is not a chatbot bolt-on or an innovation-lab experiment: the client's agentic platform sits in the production path of real customer conversations, with the reliability, latency, and safety bar that implies.
The client is engaging senior engineers who have actually shipped agentic systems — people who have felt the difference between a demo that works and an agent that survives contact with real users. The role sits within a small, senior team moving at startup pace with enterprise-scale distribution.
Responsibilities:
Designing and building production agentic workflows for customer-facing use cases — multi-turn conversation, tool calling, retrieval, escalation, and human handoff.
Owning agent orchestration end to end using LangGraph / LangChain (or equivalent frameworks), including state management, guardrails, and failure recovery.
Integrating agents with the client's core service layer: existing Java/Spring microservices, APIs, and event streams are the hands and feet of every agent built.
Building and tuning the evaluation loop — automated evals, regression suites, trace analysis, and A/B measurement of agent quality.
Working on the inference and serving layer: model routing, prompt/context management, and self-hosted serving with vLLM alongside commercial LLM APIs.
Hardening everything for a regulated, high-trust environment: PII handling, auditability, deterministic fallbacks, and observability.
Mandatory Skills:
7+ years of software engineering experience, with senior/lead-level ownership of production systems.
Proven hands-on delivery of LLM-based agentic systems in production — not coursework, not POCs. The interview covers the architecture, failure modes, and eval strategy of a shipped system.
Strong Python (primary agent/ML stack) and Java (service integration) — both are daily-use languages here.
Depth in the standard agent stack: LangChain, LangGraph, prompt/context engineering, tool/function calling, RAG pipelines.
Working knowledge of PyTorch and model fundamentals — enough to fine-tune, debug model behavior, and reason about tradeoffs.
Experience with LLM serving/inference — vLLM or comparable (TGI, TensorRT-LLM), plus commercial APIs (OpenAI, Anthropic, Bedrock/Vertex).
Major Plus
Hands-on experience with agent-builder platforms such as Sierra or Decagon — deploying, extending, or evaluating them for customer-service use cases.
Voice agents, real-time/streaming inference, or contact-center integration.
Experience in regulated industries (finance, healthcare) — model risk, compliance review, audit trails.
Eval frameworks (LangSmith, Braintrust, custom harnesses) and LLM safety/guardrail tooling.
#J-18808-Ljbffr