Staff Engineer / AI Builder - NYC

Description

The company designs AI systems for real people using them in everyday workflows, and it ships production-ready AI in 30 to 45 days.

You will be a senior technical lead on a cross-functional team. You will set AI architecture, drive model development and optimization, and work through hard integration problems.

You will join work already in flight, where reliability, security, and scale matter. You will set engineering standards and shape technical decisions on how the company's AI systems grow.

The title is Staff AI Builder and the role is based in NYC.

Responsibilities

• Design and build agentic workflows with reasoning loops, tool and function calling, and orchestration across single-agent and multi-agent setups. Decide which pattern suits each problem.
• Build and maintain RAG pipelines covering chunking, embeddings, OpenSearch vector search, re-ranking, and refresh logic for knowledge sources that change.
• Connect systems to AWS Bedrock and Agent Core, including MCP-based tools. Write tool descriptions clear enough that the orchestrator routes correctly every time.
• Write production system prompts with structured role framing, output constraints, and few-shot examples.
• Add evaluation and observability to every agent you ship. Use golden datasets, RAGAS-style metrics, and LLM-as-judge both as a runtime guardrail and offline. Trace with LangFuse, LangSmith, or an equivalent.
• Plan for LLM failure: retries with backoff and jitter, circuit breakers, fallback models, and a clear message to users when an upstream service degrades.
• Build backend services in Python and Node.js, including serverless designs on AWS Lambda and API Gateway that support agent workflows.
• Design DynamoDB single-table schemas with composite keys and transactional writes for conversation state, agent memory, and session history.
• Support asynchronous agent operations using Step Functions, SQS, and EventBridge, with strong reliability and clean error handling.
• Contribute to frontend integration points so agent features work well for end users. Write clean, well-tested code across the stack.
• Support deployment, monitoring, and production troubleshooting in AWS and Docker environments, getting senior guidance when needed.
• Bring clear technical judgment to architecture discussions, naming the real trade-offs of each pattern, such as gateway centralization versus latency or single-agent versus multi-agent design.
• Work with product, design, and delivery leads across global teams to scope and ship features end to end.
• Mentor engineers on the pod and raise the standard for agentic engineering, including AI-assisted development tools such as Claude Code.
• Own assigned features and releases end to end, including debugging, hardening, and keeping production systems reliable.

Requirements

• 6+ years of professional software engineering experience, including real time shipping GenAI or LLM systems to production, not only prototypes.
• Hands-on depth in agentic AI: reasoning loops, tool and function calling, and multi-agent orchestration, with a reasoned view on when single-agent design is better and why.
• Practical RAG skills: chunking, embeddings, vector databases such as OpenSearch, cosine similarity search, and re-ranking. You can explain how each works, not just name it.
• Experience building evaluation and observability for LLM systems, including golden datasets, LLM-as-judge, RAGAS or similar metrics, and tracing tools like LangFuse or LangSmith.
• Strong prompt engineering. You can write a production system prompt with real structure and constraints when asked.
• Hands-on work with the AWS GenAI stack: Bedrock, Agent Core, Lambda, DynamoDB with single-table design, S3, SQS, EventBridge, and Step Functions.
• Strong Python and Node.js skills, with experience building full-stack applications and RESTful APIs.
• Solid grasp of reliability patterns for LLM-backed systems, meaning retries and backoff, circuit breakers, and fallback models used together.
• Experience with Docker and cloud-native deployment.
• Comfort with ambiguous, integration-heavy problems. When challenged, you go deeper instead of staying high-level.
• Nice to have: direct experience with Amazon Bedrock Agent Core, including agent registration, MCP tool wiring, memory and session management, and gateway provisioning.
• Nice to have: experience in a regulated or high-stakes field such as education, healthcare, or financial services, where evaluation rigor and reliability have real consequences.
• Nice to have: familiarity with workflow orchestration tools such as Step Functions, Sequencer, or other state-machine systems.
• Nice to have: a informed view of LLM observability tools, based on running them in production.

Conditions

The base salary range is $165,604 to $228,491 USD. Actual pay may be higher or lower, and the range may change over time.

Pay within the range depends on factors including education, qualifications, certifications, experience, skills, seniority, geographic location, performance, shift, travel requirements, sales or revenue metrics, collective bargaining agreements, and business needs.

Benefits

• Paid time off.
• Medical, dental, and vision insurance, plus a 401(k), for eligible employees.