Principal Software Engineer, Breeze

HubSpot is an AI-powered customer platform that provides software, integrations, and resources for marketing, sales, and service. The Principal Software Engineer will serve as the technical anchor for the Breeze Assistant platform, designing scalable distributed systems and agentic AI capabilities while leading production development, reliability, trust, and cross-functional technical initiatives.

Responsibilities:

• Define the patterns for how AI agents safely interact with HubSpot's CRM, agentic primitives (skills, tools, memory, scheduled prompts), and third-party ecosystems (Slack, Google Workspace, Gong, etc.). You set the bar for how Breeze orchestrates deterministic and non-deterministic steps — choosing the right blend for trust, cost, and outcome quality
• You will extend the platform so every HubSpot team builds their highest-value deliverables into Breeze — expanding artifact types, opening the platform for internal builders, and driving toward Breeze being the default surface for customers and teams to build on
• Work closely with the Context team to ensure every Breeze interaction is powered by complete, trustworthy growth context: company and industry data, outcome history, learning loops, and customer goals. You will build the technical bridges that let context evolve from a data layer into a dynamic, active collaborator in every user's GTM work
• Act as a zero-to-one builder. You stay current with the latest in LLMs, RAG, and agentic frameworks to rapidly prototype and validate new capabilities — and you know when to ship the prototype and when to wait for production-grade infrastructure
• You ship production code. You will lead development of high-impact initiatives, moving from a prototype to a production system supporting a broad user base
• Ensure GenAI outputs are not just capable, but trustworthy. You will drive strategy around LLM evaluation, observability, guardrails, and data privacy — the technical foundation that lets customers trust Breeze enough to run real business processes on it
• Breeze is both a product and a platform. You will partner with Product, AI Research, HubSpot engineering teams, and the Ecosystem teams to ensure every HubSpot team can expose their tools and API endpoints to Breeze — because coverage is the single biggest unlock for use-case completeness
• Elevate AI fluency across the engineering org. Mentor senior engineers on prompt engineering, model selection, agentic design patterns, and the nuances of building non-deterministic systems at scale

Requirements:

• You have built and scaled high-throughput, low-latency systems with strict reliability requirements — the kind that support high concurrency and vast data scale
• Deep experience with LLM orchestration, vector databases, RAG pipelines, and multi-agent frameworks
• You understand the trade-offs between model architectures, fine-tuning strategies, and inference costs, and you make principled bets
• A background in building autonomous or semi-autonomous systems where AI takes actions, handles tool-calling, manages complex state, and recovers gracefully from failure
• You've thought hard about the interplay between deterministic and non-deterministic steps in production workflows
• You build for other builders
• You design primitives — skills, tools, APIs, agentic harnesses — that internal teams and partners can compose into use cases you haven't imagined yet
• You have a track record of building products that people love
• You advocate for the end-user in design reviews and understand the pain points of sales and marketing professionals well enough to know when a technical tradeoff harms the outcome
• You prefer a working prototype over a lengthy design doc
• You turn ambiguity into clear plans and working software, and you know how to balance the speed of AI experimentation with the rigor of production-grade systems
• You dive into a new paper, model, or framework today and can explain its strategic value — and its limitations — to the team tomorrow