Careers at Sierra AI: Teams, Pay and How to Get Hired

Who Gets Hired — and Where They Land

Sierra AI reached roughly 700 employees by July 2026 with 115-plus open roles globally, chasing $200 million in annualized revenue (per Latka) on a single product thesis: autonomous AI agents that handle customer service for enterprise brands. That focus shapes every team.

Engineering carries the heaviest headcount. The job board shows Software Engineer, Agent roles with language-specific requirements — Cantonese, Thai, and Korean speakers based in Singapore — signaling that agent deployment follows the customer's geography, not the company's headquarters. A Strategist, Agent Development role sits in Tokyo. A Product Manager, Agent Development role sits in Singapore. These aren't generic platform engineers; they build and ship agents into production for brands doing over $1 billion in annual revenue.

The go-to-market side mirrors that distribution. An Enterprise Account Executive for ANZ operates out of Sydney. Sierra hires where its customers buy. The company says it serves hundreds of customers and crossed tens of millions in annualized revenue by late 2025. That trajectory requires a sales motion that handles procurement at global enterprises, not just demos to developers.

Inside engineering, a six-person AI acceleration team runs experiments on Sierra's own agent stack. Formed in early 2026 after engineers using git worktrees, Claude Code, and Codex reported 5× productivity gains on some tasks, Sierra found, the team maintains internal tooling — Pinecone, MCP Gateway, Agency — through named engineers Allen Chen, Mihai Parparita, and Rohith Ravi. Their work feeds back into the agent product customers use.

Early-career hires enter a structured program that places them directly with customers from day one. The careers page describes the track as "working directly with big-name brands to build and launch AI agents that impact millions of consumers, all with the mentorship of amazing teammates." New hires don't sit in a training sandbox; they ship to production under supervision.

What's missing from the public view is a traditional research science org. Sierra doesn't publish papers on foundation model architecture. Its engineers apply frontier models — Claude, GPT-4, Codex — to the orchestration, tooling, and evaluation layers that make agents reliable in production. The hiring profile reflects that: more systems engineers and product-minded builders, fewer pure ML researchers.

The org chart, in practice, reads like a professional services firm that productized its delivery layer. Every team either builds the agent platform, deploys it for a specific customer, or sells the next deployment. Headcount grows where the next contract lands.

This guide maps the known hiring picture at Sierra AI: compensation bands from internal data, the interview process as disclosed by the company, physical work environments at its sites, and the candidate traits highlighted in employee accounts.

What the Offer Looks Like

Sierra's compensation sits at the upper end of the frontier-AI market. Compensation data from 148 salaried postings shows a typical band of $160,000–$403,000 with a median of $321,000 — higher than third-party aggregates because it weights toward senior engineering and agent-development roles, which dominate open headcount.

Third-party sources bracket the same range from different angles. Levels.fyi records a low-end of $93,132 for a Recruiter and a high-end of $460,000 for a Software Engineer. Recruiting from Scratch analyzed 67 public postings from 2025–2026 and found a median of $250,000 with a middle-50% band of $221,000–$291,000. Jobs by Culture calculated a median base salary of $250,000 from nine H-1B filings. An AshbyHQ posting lists $180,000–$390,000 plus equity. The consistency across independent cuts suggests the $250,000 median is durable, while the board's $321,000 median reflects the current hiring mix.

*USD conversions at mid-2025 spot rates; actual offers denominated in local currency.

The Singapore engineering bands are notable: a S$200,000 spread (roughly $148,000) between floor and ceiling on the same title signals that Sierra prices for demonstrated agent-shipping velocity, not years of experience. The Tokyo strategist role sits lower in absolute dollars but is competitive for the Japanese market. The Sydney enterprise AE band is tight — A$80,000 wide, reflecting a quota-carrying role with less variable equity upside.

Equity is standard across technical and GTM roles; the AshbyHQ posting explicitly notes "Offers Equity." Sierra's careers page lists top-tier health and life insurance, a 401(k) with company match, generous parental leave, fertility benefits, and flexible time off. An annual stipend — usable for "a new monitor, childcare, or a phone plan" — sits on top of base and equity. For candidates outside the San Francisco Bay Area, the company confirms relocation assistance on its early-career FAQ. Visa sponsorship is offered case by case: "we do sponsor visas… we'll do our best to support your visa process."

No public data breaks out refresh grants or sign-on structures. The H-1B median of $250,000 base aligns with the board's lower quartile, suggesting visa-dependent hires may enter near the band floor and scale quickly. For a candidate comparing offers, the actionable takeaway: negotiate from the role-specific band above, not the company-wide median. The Singapore agent-engineer ceiling of S$495,000 ($366,000) is the highest posted cash figure in the current dataset; the Tokyo strategist floor of ¥20,000,000 ($130,000) is the lowest.

How the Loop Runs Now

Sierra's interview process underwent a deliberate redesign that mirrors the company's own product shift. The old loop (two coding interviews, an algorithms round, system design, and culture fit) optimized for mechanics: typing syntax, recalling algorithm details, stitching frameworks together. That format produced signal, but the wrong kind. In debriefs, hiring managers found themselves leaning on referrals and prior experience because the interviews didn't reveal how candidates actually build with AI. The company acknowledged the gap publicly on its engineering blog, then replaced the phone screen and coding rounds with an AI-native onsite built around three attributes: representative of daily work, high signal on strengths and gaps, and a positive candidate experience.

The current funnel typically runs three to five weeks. It opens with a 30-to-45-minute recruiter screen that still covers data structures and algorithms. LeetCode-style preparation remains sufficient here, according to candidate reports. From there, candidates receive a paid take-home assignment: build a customer service agent. Several days later, the onsite loop (often in San Francisco) centers on that artifact. The new onsite has three phases. First, a planning session where the candidate drives product ideation in their domain while interviewers probe and sharpen the concept. Second, a two-hour build window: the candidate brings the idea to life using whatever AI tooling and frameworks they choose, with explicit permission to cut scope and skip boilerplate. Third, a review where the candidate demos the working result, debates key product flows and trade-offs, walks through code for technical judgment (data model, abstractions, extensibility) and explains how they used AI along the way. Paul Buchheit's line frames the bar: if it's great, it doesn't have to be good.

Sierra also replaced the traditional coding phone screen with a system design interview, reasoning that vibe-coding an app is easy; getting it into production at scale is the harder, more relevant problem. A debugging interview is now in pilot: candidates receive a medium-sized codebase and a draft PR introducing a cross-cutting feature, then review and improve it using coding agents. The level of AI assistance permitted in that round remains unsettled as new models zero-shot many fixes. For infrastructure candidates, the format is amended to better capture vertically integrated product work.

Across every stage, evaluation criteria map to five stated values: Trust, Customer Obsession, Craftsmanship, Intensity, and Family. Interviewers press hard on the take-home discussion, not for a polished demo, but for the reasoning behind product choices, failure modes, and trade-offs in agent behavior. One candidate described being "grilled hard" on the customer service agent they built. The debugging round surfaces basic defects (incorrect strings, off-by-one loops), not to test trickiness but to see whether candidates stay precise under pressure and spot defects quickly. A past-project deep dive demands frontier-level engineering experience: candidates must own a genuinely hard decision, defend the trade-offs, articulate what broke, and explain what they would change. Vague ownership collapses under follow-up.

Preparation guidance from the company and candidate community converges on three priorities. Fluency in TypeScript or Python and at least one LLM API, streaming, tool calling, token limits, well enough to reason about failure modes on the spot. Deep understanding of agent orchestration, tool-use, and production reliability: validation, retries, fallbacks, human handoff, prompt-injection containment. And familiarity with Bret Taylor's public talks and Sierra's engineering posts so answers use the company's vocabulary. Candidates who advance tend to demonstrate high agency, a builder's mindset, and clear justification for every design decision. The process is generally well-organized and fast, though a minority report disengaged interviewers or an intense motivation screen. As one accepted candidate put it: the interview tested what you can actually do, not what you can memorize.

Where the Desks Are

Sierra AI operates from a distributed footprint that reflects its dual character: a San Francisco–anchored product and engineering core, a growing go-to-market presence in major commercial hubs, and an international layer that mirrors where its enterprise customers sit. The company describes San Francisco as "a core Sierra AI office location" providing "access to Silicon Valley's innovation ecosystem and leading tech talent." That office houses product, engineering, and design teams building the conversational AI platform, including agent development, SDK tooling, and the Ghostwriter agent-building agent shipped to customers.

Beyond the Bay Area, research identifies three additional U.S. sites tied to Sierra's commercial motion. An East Coast office (the research names the function but not the city) "supports business development and client engagement activities in the major East Coast business hub." Atlanta appears explicitly: "The Atlanta office enhances reach in the Southeast's growing tech ecosystem and business community." London rounds out the confirmed locations: "London serves as a strategic international location, supporting operations and partnerships in Europe."

First-party board data from Zero G Talent adds signal on where roles are actually posted. Recent listings include a Strategist, Agent Development role in Tokyo; multiple Software Engineer, Agent positions in Singapore (tagged for Cantonese, Thai, and Korean language support); a PM in agent development in Singapore; and an Enterprise Account Executive (ANZ) in Sydney. These postings suggest either established hubs or active hiring pipelines in Tokyo, Singapore, and Sydney, locations not mentioned in the company's own office directory but visible in live recruiting data.

The research draws a separate line for Sierra Space, a distinct entity that shares the Sierra Technologies, Inc. corporate umbrella but operates its own facilities. "The primary Sierra Technologies, Inc. headquarters, located in Broomfield, Colorado, offers easy access to both Boulder and Denver and is situated near FlatIron Crossing, known for its shopping and dining. Serving as the Sierra Space headquarters, this location is central to major company initiatives and strategic operations." Sierra Space board postings confirm active hiring in Broomfield, Louisville, and Centennial, Colorado, all within the Denver metro area. Candidates interviewing for Sierra AI roles should not conflate the two; the Colorado sites belong to a sibling business focused on spacecraft, payloads, and space-superiority programs, not the conversational AI platform.

Physical workspace details, including square footage, lab build-out, collaboration spaces, and commuter amenities, are not disclosed. The company's public materials emphasize outcomes (agent deployment, customer lifetime value, outcome-based pricing) over facility tours. What the research supports is a geographic strategy: engineering and product gravity in San Francisco; revenue and partnership coverage in Atlanta, the unnamed East Coast hub, London, and across APAC (Tokyo, Singapore, Sydney); and a separate aerospace campus in Colorado. Candidates should ask recruiters which site a given role anchors to, whether relocation is expected, and how much cross-site collaboration the team's workflow demands, especially for roles at the product–customer boundary where travel to enterprise accounts is routine.

The Profile That Sticks

Sierra's values, including Trust, Customer Obsession, Craftsmanship, Competitive Intensity, and Family, read like a standard corporate poster until you see how they filter for them. The company is explicit: "Working side by side helps us move faster and solve harder problems." That in-person mandate in San Francisco isn't negotiable, and it selects for people who treat proximity as a productivity tool, not a policy to tolerate.

The engineering team skews senior. Reviewers consistently describe colleagues as "the sharpest people they've ever worked with." That concentration of experience means less formal mentoring infrastructure but more opportunity to work alongside people who have already built systems at massive scale, including Google Maps, Salesforce, and Google Labs. For engineers who learn by osmosis, that access is the real compensation. Two deeply technical co-founders mean engineers have genuine influence over product direction; product managers don't hand down specs. The ship-fast culture produces short cycles from idea to production, but the enterprise context demands that speed doesn't compromise reliability. Agents handle billions of real customer interactions, processing insurance claims for Cigna, scheduling service for Rivian owners, managing returns for ADT. The feedback loop is concrete and fast.

Ambiguity tolerance is non-optional. The company doubled its headcount in 2025 and closed three acquisitions in early 2026 alone, including Receptive AI for voice technology, Opera Tech for Tokyo-based enterprise deployment, and Fragment for workflow orchestration. Roles evolve weekly. Processes are being written in real time. If you need a clearly defined lane with stable handoffs, this isn't it. The people who stay treat organizational chaos as raw material.

Competitive Intensity appears as a value alongside Trust and Craftsmanship, an unusual pairing that signals the actual culture. This is a place that celebrates winning deals, shipping fast, and outperforming. The verified work-life balance score sits at 3.5 out of 5, well below the 4.6 on Glassdoor; reviewers explicitly warn "don't expect this to be a chill 10–4 tech job." Startup intensity with enterprise-grade stakes. The equity is backed by $150M+ ARR at a $15.8B valuation, real revenue, not future projections — Sierra's data shows — so the financial upside tracks the operational pressure.

Founder-caliber mentorship exists but isn't structured. Bret Taylor and Clay Bavor are present in the building. Working under them is a career accelerant for people who can extract signal from access without needing a formal program. The flexible personal stipend, usable for whatever an employee decides improves their work life, reflects a trust-first default that extends to how teams operate. Autonomy is high; guardrails are few.

The profile that emerges: senior engineers who want to own ambiguous problems end-to-end, product people who prefer shipping to spec'ing, sales and operations talent motivated by visible customer impact over process perfection. All of them comfortable in person, five days a week, in a culture that measures output in production agents handling real transactions. The 157 open roles across engineering, product, sales, and operations suggest the filter is working, or at least that the company knows exactly who it's building for.

The same six-person AI acceleration team that formed after engineers reported 5× gains using Claude Code and Codex now maintains the internal tooling that ships to customers. Chen, Parparita, and Ravi aren't building demos; they're building the infrastructure that lets every other team move faster. That loop, from experiment to production to the next experiment, is the only org chart that matters.

Working in frontier tech? Zero G Talent tracks the openings: see every open Sierra AI role, browse frontier tech jobs, openings at Sierra Space, and the people building the field.