Principal AI Engineer

About the role

Principal AI Engineer

NYC (Hybrid 3 Days)

Full-Time

Seeking a Staff/Principal AI Engineer to lead enterprise-scale Agentic AI solutions for Fortune 500 clients. Must have deep hands-on experience building and deploying production-grade LLM applications, autonomous agents, RAG systems, MCP servers, vector databases, and distributed backend systems.

Roles & Responsibilities

Design and build agentic systems: Lead the architecture and implementation of tool-calling agents that combine retrieval, structured reasoning, and secure action execution with least-privilege access.

Productionize LLM applications: Build retrieval pipelines, prompt synthesis, response validation, and self-correction loops, backed by rigorous evaluation.

Own the full stack: Deliver the data pipelines, backend services, distributed compute, and orchestration layer that agentic systems depend on — not only the model invocation.

Engineer for reliability and governance: Build validator models, adversarial test suites, and policy checks; enforce deterministic fallbacks and rollback strategies; instrument continuous evaluation.

Optimize for cost and latency: Drive measurable improvements in token efficiency, response time, and unit economics against defined SLOs.

Codebase ownership: Build, maintain, and review high-quality Python and SQL, with an emphasis on reusable components, scalability, and performance.

Cloud integration: Deploy AI applications on AWS, Azure, or GCP with optimized resource usage and robust CI/CD.

Cross-functional collaboration: Partner with product owners, data scientists, and business SMEs to define requirements and deliver impactful AI products.

Mentoring and technical leadership: Set engineering standards and share knowledge across the team, raising the bar on AI and software engineering practice.

Required :

8–14 years of software engineering experience, with strong hands-on large-scale Python

Working depth in at least one systems or backend language — Go, Rust, Java, or C/C++ — and the judgment to know when to reach for it

Strong data structures and algorithms.

Strong understanding of APIs, microservices, and system design

Hands-on experience building and operating data pipelines and production-grade distributed systems.

Agentic AI and LLMs

2+ years of hands-on LLM engineering, with at least couple agentic system you designed and took to production

Production experience with agent frameworks — LangGraph, Google ADK, CrewAI, Claude Agent SDK, or equivalent — and the fluency to move between them as the ecosystem evolves

Experience building MCP (Model Context Protocol) servers and tool-calling interfaces

RAG from first principles: chunking strategy, embeddings, vector and hybrid retrieval, reranking, and response validation

Strong experience with vector databases (Milvus, Pinecone, Weaviate, FAISS, etc. or cloud equivalents)

Design of guardrails and reliability patterns — validators, policy checks, self-correction loops, deterministic fallbacks, circuit breakers, and rollback paths

Optimization

Deep familiarity with token optimization and context-window management — context shaping, pruning, and compaction

Latency and cost optimization through caching, model routing, batching, streaming, and parallel tool calls

Performance testing and tuning systems against defined SLOs

Evaluation

Experience building evaluation frameworks for LLM systems — offline eval sets, continuous online evaluation, and regression detection

Instrumentation and traceability suitable for regulated enterprise environments using tools like LangSmith, Langfuse, etc.