About this role.
Help Scout is hiring a Staff Product Engineer for its Agents team to deliver production-grade, customer-facing AI capabilities across the full stack. The role has a frontend emphasis, requiring strong TypeScript, JavaScript, and React craft alongside ownership of LLM-backed retrieval, embeddings, and agentic tool-calling systems. The engineer will build evaluation, guardrail, observability, reliability, and graceful-degradation mechanisms for AI features, while measuring customer and product outcomes after launch. This is a highly autonomous remote role that involves close collaboration with product managers, designers, customers, and fellow engineers. The position is U.S.-based and offers a disclosed annual salary range of $187,000 to $214,000 USD.
Role DNA
A quick view of the complexity, pace, ownership and collaboration implied by the job description.
Job Complexity
5/5Pace & Pressure
5/5Autonomy Level
5/5Communication Load
5/5Salary analysis
Estimated compensation compared with the broader US market for similar roles.
Core skills
Skills and capabilities most closely associated with this opportunity.
Sample interview questions
I would explain the customer problem, the system design, and the quality and business metrics defined before launch. I would cover the representative test set, automated and human evaluation methods, monitoring, and the iteration decisions made after observing real customer behavior.
I would begin with permission-aware retrieval, source-quality controls, chunking and ranking experiments, and clear grounding of responses in retrieved context. I would add confidence signals, citation or source visibility where useful, fallback behavior when retrieval is weak, and evaluations for factuality, relevance, latency, cost, and agent acceptance.
I would prioritize responsive streaming states, cancellation and retry behavior, accessible status updates, clear distinctions between generated and user-authored content, and safe editing controls. I would also instrument interaction points to understand completion rates, corrections, abandonment, and whether the feature builds user trust.
I would scope an initial customer outcome and establish non-negotiable safeguards such as observability, error handling, evaluation coverage, and rollback paths. I would ship incrementally behind feature controls, monitor quality and operational metrics, collect customer feedback, and use the evidence to improve or reconsider the solution.
I would describe the original hypothesis, the customer evidence or metrics that challenged it, and how I partnered with product and design to adjust the implementation. The strongest example would demonstrate ownership after launch, transparent communication about what did not work, and a measurable improvement from the revised approach.
Who we are
Help Scout started in 2011 with a simple premise: support software should treat customers like people, not tickets. 3+ billion conversations later, more than 10,000 companies use Help Scout to grow relationships with customers of all types, from clients to patients and beyond.
Now we’re in a new chapter: not only building AI that gives teams more space to focus on the customer experience, but also reimagining how we operate in an AI-native world. That means shipping and learning faster, without compromising on craft. Great work here carries a strong opinion and sweats the details, regardless of the tools we use to drive it.
We’ve been a fully remote team since day one, with 120+ teammates now all over the world. It’s a culture where you’ll be trusted to take ownership, stay curious, and raise the quality of work around you. It comes with clear expectations and a team that will push you to do some of the best work of your career.
If that kind of challenge gives you energy, we’d love to meet you.
About this role
This role sits on our Agents team — the group building the AI-powered experiences our customers use every day. It’s where large language models stop being a demo and become a dependable part of the product: answers, drafts, and assistive features that have to be fast, trustworthy, and genuinely useful in front of real customers.
We’re looking for an engineer who can build that end-to-end — from the retrieval and agentic plumbing behind a feature to the polished, fast, accessible interface a support agent actually touches. You’ll have a frontend center of gravity (strong JavaScript/TypeScript and real front-end craft) while operating as a full-stack product engineer who owns the whole thing.
How we work
Engineers at Help Scout own their work end-to-end — from problem definition through production, and from the customer signal that shaped the bet to the dashboard that tells you whether it landed. You’re full-stack capable, customer-fluent, and genuinely fast with AI tools. You ship the whole thing, not just your layer.
We develop in Kotlin, Java, TypeScript, React, and PHP.
We leverage Elasticsearch, ClickHouse, Kafka, Flink, RabbitMQ, MySQL, and PostgreSQL — plus vector search and LLM tooling for our AI features.
We’re heavy users of the AWS ecosystem, Slack, GitHub, and Linear.
We use AI tools — Cursor, Claude Code, and a growing set of internal tools — every day. We treat our AI workflow as a product we’re shaping together, and we expect you to help shape it.
Teams are small and project-based. You might be 3 engineers and a PM on one project, 2 engineers and a designer on the next, or holding more of the product thinking yourself. You’re expected to operate well across different compositions and ship end-to-end.
What you’ll do
Ship customer-facing AI features across the full stack. Build the experience and what’s behind it — LLM-backed flows, retrieval (RAG) and embeddings, and agentic tool-calling — through to a fast, accessible, well-crafted UI. Most engineers here have a strength; yours leans frontend, and you take initiative across every layer.
Treat AI output as something you verify, not assume. Build evaluations, guardrails, and monitoring into AI features so quality is measurable and holds up in production. Know when to trust a model’s output and when to push back — and design the system so the product degrades gracefully when a model or provider misbehaves.
Sweat the front-end craft. The interface is where AI either earns or loses customer trust. You care about responsiveness, streaming/loading states, accessibility, and the small details that make an assistive feature feel reliable rather than gimmicky.
Stay close to customers. Join customer calls when more context is needed to design the right solution, participate in a support escalation rotation, and watch session recordings on the features you own. Bring that signal back into the work.
Own outcomes, not just output. Before a project starts, you and a PM agree on a specific outcome metric the work is trying to move. You instrument it, and after it ships you’re watching the dashboard and talking to customers — deciding what to do next, not just closing a Linear issue.
Co-author the solution. PMs own strategy and sequencing; you bring the judgment and craft for how solutions actually get built, and you’ll often shape the product thinking too. The best work happens when engineers, designers, and PMs support each other.
Own production readiness from the start. Automation, reliability, monitoring, alerting, and logging aren’t afterthoughts — they’re part of how you ship. The work continues after merge.
Use AI tools every day as part of your craft. We expect fluency with Cursor, Claude Code, or similar. If you saw a way to make the team’s AI workflow better tomorrow, you’d say so.
Help us hire. From time to time, partner with our Talent team to interview future teammates — one of the highest-leverage things any engineer here does.
About you
You’re a strong, full-stack-capable engineer with a frontend center of gravity. You have substantial JavaScript/TypeScript and React experience and real front-end craft, and you take initiative across the stack and ship end-to-end without waiting for someone to own the other half.
You’ve built real things with LLMs — not just prototypes. You’re comfortable with retrieval (RAG), embeddings, prompt/context engineering, and agentic/tool-calling patterns, and you have a point of view on what makes AI features actually reliable in production.
You hold a high bar for AI quality. You think in terms of evaluation, verification, and guardrails; you measure whether an AI feature works rather than assuming it does; and you know that more AI-generated code is not the same as more value.
You’re genuinely fluent with AI coding tools and treat them as part of your craft. You move fast without sacrificing judgment, and you have a point of view on where these tools are heading.
You’re customer-fluent. You actively seek out customer signal because it’s how you stay connected to whether your work is actually solving the problem.
You own outcomes. You measure your work by whether it landed — and you’re comfortable saying “this didn’t move what we hoped, here’s what we want to try next,” then following through.
You take real ownership of the full development lifecycle — automation, reliability, resilience, monitoring, alerting, and logging built in from the start. You stay with what you ship until the metric moves and the customer is better off.
You communicate clearly in writing and in conversation. Help Scout is fully remote and writing is the medium of most decisions. You give and receive direct feedback, and you see code review and pairing as real chances to teach and learn.
Nice to have
Experience making LLM features production-grade: latency/cost tuning, fallbacks and circuit breakers across providers, moderation/safety, and handling sensitive data responsibly.
Familiarity with evaluation/observability tooling for AI (LLM-as-judge, test sets, tracing) and the discipline of building representative eval sets.
Design sensibility — comfort partnering closely with designers and elevating the craft of an interface, not just implementing a spec.
Experience in customer support, productivity, or other tools where trust and reliability are the product.
Company values
Happy to Help
Help is in our first name! We show up for each other — not out of obligation, but because we’re invested in the team’s collective success. We share knowledge freely, lead with generosity, and practice empathy with our teammates, customers, and community.
Craft over Convention
Our success relies on the quality and craft of the work we put into the world. The status quo simply won’t work. So we insist on narrow focus, sweating every detail, and relentless pursuit of customer delight.
Progress not Perfection
Achieving our true potential — collectively and individually — requires constant progress and forward momentum. By creating a culture of curiosity and openness, we aim to create a safe space for mistakes, the ability to identify them quickly, and use them to get better.
Own the Outcome
Own the outcome means taking full responsibility for the results of your work, decisions, and contributions. It reflects a mindset of accountability, proactiveness, and follow-through. If you “own the outcome,” you don’t just complete tasks, you ensure your work leads to meaningful results, and take initiative to solve problems rather than passing them along.
Read more about how we define, share, and live these values here.
Benefits And Perks
Competitive salary and an internal, transparent salary formula based on market data
Flexible time off – you choose the holidays and vacations that make sense for you
12 weeks of fully paid parental leave for all new parents, including adoption and foster care
A home office stipend to help you get set up and productive
A co-working stipend up to $300 a month if you choose to work out of your house
A yearly professional development stipend of $1,800 to help you grow in your craft
If you’re in the U.S. or Canada, we offer top tier health insurance for you and your dependents.
Hiring Guidelines:
All roles at Help Scout are fully remote. Some positions have specific location requirements. For roles outside the U.S. and Canada, we work with teammates as independent contractors under country-specific agreements.
We do not sponsor visas. Teammates must already be authorized to work from their home country.
We are an equal opportunity employer and are committed to building a company that embraces and celebrates diversity and inclusion. We do not discriminate on the basis of race, religion, color, national origin, gender, gender identity or expression, sexual orientation, age, marital status, veteran status, or disability status. We have read the studies and understand that diverse teams build better products, bring more perspective to the table, contribute to a company’s financial success, and help foster a more inclusive environment for all employees, but the bottom line is that it’s the right thing to do.
To provide you with the best experience, we can support you with accommodations or adjustments at any stage of the recruitment process. Simply inform our recruitment team during your conversation with them.
A note on the use of AI in our interview process
While we understand that AI is an important tool in a modern skillset, we ask that you do not utilize AI assistants live on our interviews. Of course, use any tools you need to prepare! But we’d like for you to have a real human-to-human conversation with our team.
On our end, we utilize a platform called Brighthire as a notetaker and review tool to ensure a fair and transparent hiring process.
Beware of scammers! All legitimate communication from Help Scout regarding your application will come directly from an @helpscout.com email address and through our applicant tracking system, Ashby. We will never ask for personal financial information or conduct interviews outside of these official channels.
Upload your resume now
- To unlock remote work opportunities and be discovered by global employers.
This job listing has been manually reviewed by the Jobicy Trust & Safety Team for compliance with our posting guidelines, including verification of the company's legitimacy, accuracy of job details, clarity of remote work policy, and absence of misleading or fraudulent content.
