Principal Software Engineer, Breeze

Job description

POS-33518 About the Team Breeze Assistant is HubSpot's AI-powered gateway to value — the conversational front door through which every HubSpot customer starts their work, gets outcomes, and grows with the product. Our mission is to make Breeze the fastest, best way to achieve sales and marketing outcomes in HubSpot: from a new customer's very first session through the seasoned marketer orchestrating complex multi-step GTM workflows. We significantly increased both CSAT and weekly active penetration by building a foundation-first, tightly-scoped system. We are now running that same playbook at scale — shipping a full agentic platform that every HubSpot team builds on. The next leg of the journey is larger: Breeze becomes the place a customer can get any GTM task done, powered by complete customer context, agentic everboarding, and a platform mature enough that partners and internal teams build their most important use cases on top of it. If a marketer or seller can get their real work done — and done better — inside Breeze, they stay in HubSpot, come back, and grow. We build so they never have to leave. About the Role We are seeking a Principal Software Engineer to be the technical anchor for the Breeze Assistant platform. This role sits at the intersection of robust distributed systems, bleeding-edge Generative AI, and real customer outcomes. You are not just following a roadmap — you are helping define it. You will shape how Breeze moves from a highly capable assistant to a trusted, proactive agent: one that delivers durable, shareable artifacts grounded in CRM data; dynamically generates the UI a user needs for the task at hand; orchestrates multi-step workflows across HubSpot and third-party tools; and learns from every interaction to get better. The scale is real — millions of users, petabytes of context, and a platform that every HubSpot engineering team depends on. Key Expectations • Agentic Platform Architecture: 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. • Building Breakthrough Surfaces: Breeze already delivers a significant volume of artifacts across reports, decks, and pages. 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. • Complete Customer Context: 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. • Rapid GenAI Experimentation: 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. • Hands-on, High-Leverage Contributor: 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. • System Reliability & Trust: 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. • Cross-Functional Technical Leadership: 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. • Technical Mentorship & Multiplier: 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. What You Bring • Proven Expertise in Distributed Systems: You have built and scaled high-throughput, low-latency systems with strict reliability requirements — the kind that support high concurrency and vast data scale. • GenAI Proficiency: 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. • Agentic System Design: 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. • Platform Thinking: 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. • User-First Mindset: 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. • Bias for Action: 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. • Rapid Domain Fluency: The AI landscape changes weekly. You dive into a new paper, model, or framework toda