The hiring map: where Tavus places its bets
Of the 10 to 14 roles Tavus lists across its careers page and aggregators as of September 2026, fewer than half sit in engineering — the rest belong to product, design, sales, and business development, a split that contradicts the company's research-lab framing. EngRadar counts 14 active roles, DreamworkHQ sees 10, and the company's own Ashby board shows a similar range — but the functional shape holds across all three. Engineering, product, and design sit at the core. Go-to-market functions, such as sales, marketing, customer success, and business development, occupy nearly as many slots. Research carries a smaller but distinct presence. The first-party Zero G Talent board, which ingests postings directly from Tavus's ATS, confirms the pattern: recent listings include a Senior Software Engineer (Infrastructure), a Forward Deployed Engineer, a Marketer focused on brand and product storytelling, a Customer Engineer, a Business Development Representative, and a "Vibe Growth Marketer." Three engineering-adjacent roles, three commercial ones.
The Ashby board groups roles into four buckets: Customer Success (2), Engineering/Product/Design combined (8), GTM (3), and Sales (1). EngRadar breaks it differently — engineering 7, design 2, product 2, sales 2, marketing 1. DreamworkHQ reports Engineering 3, Customer Success 2, Design 2, Research 2, Product 1. FastAI Jobs lists AI/ML Research (2), Customer Success & Solutions (2), Data & Analytics (1), Design (2), Marketing & Community (2), Product (1), Sales & Partnerships (2), Software Engineering (2). No two sources agree on the exact count, and some categorize the same role differently — Forward Deployed Engineer could sit in engineering or customer success depending on the taxonomy. But the aggregate picture is consistent: Tavus is not a pure research shop hoarding PhDs. It is a product company staffing the full loop from model to market.
Seniority skews heavily experienced. DreamworkHQ's sample shows 70 percent of open roles at senior level, 30 percent mid. EngRadar lists 13 roles with unspecified seniority and just one explicitly senior. The first-party board data shows salary bands that align with senior hiring — infrastructure engineering at $160k–$250k, forward deployed at $140k–$200k, marketing at $120k–$150k. The board's overall band runs $80k–$235k with a $175k median across four salaried roles posted directly.
Geography reinforces the commercial tilt. Twelve of the Ashby listings are San Francisco-based; three are remote-eligible. DreamworkHQ puts seven roles in San Francisco and three remote. EngRadar reports 57 percent remote-eligible. The company's own careers page emphasizes San Francisco presence with semi-annual full-team retreats, but the remote fraction has grown since the Series B close in November 2025 — a $40 million round that brought total funding to $64.2 million and likely triggered the current hiring wave.
What the staffing map reveals is a company that has moved past the "does the model work?" phase. Tavus's proprietary stack, including Phoenix-4, Sparrow-1, Raven-1, and the newly released Phoenix-4.5 real-time rendering model, is already serving over 100,000 developers and enterprises across recruiting, sales, education, and customer service. The hiring priority now is making that technology usable, sellable, and supportable at scale. Engineering headcount stays lean because the core research team is small and the product surface is API-first. The expansion goes to the functions that turn API calls into revenue: product managers who shape developer experience, designers who make the studio intuitive, forward-deployed and customer engineers who unblock integration, marketers who translate technical capability into use cases, and business development reps who open enterprise doors.
The research lab framing remains real — Tavus still hires AI researchers and publishes on multimodal generation. But the headcount allocation says the company's next milestone is commercial traction, not another paper.
Compensation: what the bands show
Tavus's compensation structure reveals a company building for commercial scale, not research prestige. The board's first-party data, four salaried roles posted with bands, shows a consistent range of $80,000 to $235,000 with a median of $175,000. That band aligns with the lean-engineering, heavy-GTM hiring mix documented above: the company pays market rates for product, design, and sales talent while keeping the engineering headcount small enough that each hire carries disproportionate leverage.
| Role | Location | Salary Band (USD/year) | Employment Type |
|---|---|---|---|
| Senior Software Engineer (Infrastructure) | San Francisco, CA / Remote (US) | $160,000 – $250,000 | Full-time |
| Forward Deployed Engineer | San Francisco, CA | $140,000 – $200,000 | Full-time |
| Marketer (Brand, Product, Storytelling) | San Francisco, CA | $120,000 – $150,000 | Full-time |
| Customer Engineer | San Francisco | $30 – $40 / hour | Contract |
| Business Development Representative | San Francisco / Remote (US) | Not disclosed | Full-time |
| Vibe Growth Marketer | San Francisco | Not disclosed | Full-time |
Source: Zero G Talent board postings (live at time of ingest).
The spread is telling. The top of band for a senior infrastructure engineer, $250,000, sits well below the $703,500 median total compensation that Levels.fyi reports for Tavus software engineers as of September 2026. That figure appears to be an outlier driven by a tiny sample (Levels.fyi shows a single role family) and likely reflects an exceptional equity event rather than the standing offer. The board's band, by contrast, is drawn from active requisitions and matches the $82,902–$106,148 average range Salary.com estimates for Tavus Inc overall. In a frontier-AI market where compensation benchmarks shift quarterly, the board's $80k–$235k band is the operable number for candidates negotiating today.
Equity mechanics follow the pattern a comparable frontier-AI voice company has made explicit in its public compensation-hiring brief: ElevenLabs states that "employees should be genuine owners in the company and invested in our long-term ambition." Tavus does not publish grant sizes or strike prices, and the board data does not include equity columns — a deliberate omission. What the research shows across the sector is a standard Min–Mid–Max band architecture where the midpoint tracks the 50th percentile of market data, the minimum the 25th, and the maximum the 75th. Grants are typically expressed as a percentage of the outstanding fully-diluted pool and vest over four years with a one-year cliff. Candidates should ask where a given role sits in the job-architecture leveling (IC1/IC2/IC3 or manager track) and what percentile of the band the offer targets. ElevenLabs' compensation-philosophy brief emphasizes "design systems that introduce rigor without slowing the company down" — meaning bands are guardrails, not straitjackets, and exceptions for critical hires are documented and governed.
Pay transparency laws in California and other jurisdictions now require ranges in postings, which is why the board data exists at all. But Tavus's bands also serve the internal equity goals the HR literature stresses: preventing inversions where a junior hire's package overlaps a senior's, speeding offer cycles, and giving managers a defensible framework for performance reviews. The board's median of $175,000 across four salaried roles suggests the company calibrates to the 50th–60th percentile of Bay Area market data for each function — competitive but not speculative. For a team translating AI research into commercial products, that discipline matters: every dollar allocated to cash compensation is a dollar not spent on compute, data, or the next model release.
Inside the interview loop
Tavus runs a compact, multi-gate process that reflects its size, roughly 20 people per Levels.fyi and BuiltIn, and its product: the company uses its own AI video agents to screen candidates at scale. The pipeline typically spans three to four interviews plus a technical task, a November 2025 Glassdoor review said. The first gate is often an AI-led conversation, not a human recruiter.
The AI screen: Lucas, the PAL interviewer
Tavus deploys a predefined "PAL" (Persona Abstraction Layer) named Lucas to conduct first-round case interviews. Lucas is described in the company's own documentation as a "seasoned, professional AI interviewer designed to do so with a structured yet approachable style, focusing on assessing communication and problem-solving skills." The agent runs a fixed 30-minute structure: introduction and background questions (5 minutes), case presentation (3 minutes), candidate-led analysis with guidance as needed (15 minutes), candidate questions (5 minutes), and wrap-up with next steps (2 minutes). The system scores every applicant on the same rubric and hands the hiring team a ranked list — "consistent, ranked results" is how Tavus markets the feature to its own customers.
Candidates should know they are talking to a machine. Gartner's October 2025 guidance, cited on Tavus's site, urges recruiting leaders to "clarify how they use AI in hiring and let candidates opt out of AI-led interview formats," noting that "unclear expectations are a recurring source of candidate frustration." Tavus's own blog acknowledges this: "Candidates also want to know when they're interacting with an AI." If you prefer a human screen, ask the recruiter before the first round.
Recruiter phone screen: three filters in one call
If the AI screen advances you, or if the role bypasses it, a recruiter phone screen follows. Glassdoor's November 2025 account confirms a recruiter screen occurs early in the process. Industry-standard recruiter screens serve three purposes: narrow the applicant pool, verify basic requirements (years of experience, core stack, location eligibility), and assess culture and personality fit. A widely cited recruiting walkthrough breaks these down explicitly and gives tactical advice that signals what hiring teams expect: take the call in a quiet place at a desk, not walking the dog, not from a coffee shop. Answer with "Hi, this is [your name]" and smile; "you can hear a smile." Keep your résumé, the job description, and the interviewers' names printed in front of you, because the recruiter can't see your notes.
First human rounds: general fit, then the case
Glassdoor's November 2025 account describes the first two human interviews as "general fit ones that weren't technical but just going over the role." This aligns with the recruiter's gate: the team confirms motivation and alignment before investing technical time. The third round brings the hiring manager. A fourth round may follow for some roles.
For technical and product roles, expect a technical task; the Glassdoor reviewer notes it explicitly. Design candidates face a portfolio review. A LinkedIn post from a Tavus hiring manager (hiring two designers as of the post date) said they see "a TON of fantastic Designers' work" and that candidates often ask for "a 15-min call, advice, or referrals." The same post links to "Tavus Design Hiring: 5 Tips to Stand Out in a Portfolio", a signal that portfolio narrative, not just pixel polish, carries weight.
What a strong application looks like
- Tailor to the job description. Tavus's AI application review (used internally and sold to customers) "presents a ranked list of candidates based on job fit, with scoring derived from auto-generated criteria within the job description," per Select Software Reviews' 2026 ATS test. Mirror the JD's language in your résumé and cover letter.
- Show product intuition. ** Tavus builds AI video agents that "read tone and expression the way text tools never could"; Sparrow leads TurnBench for conversational turn-taking, Phoenix renders at 134 ms latency, Raven ingests voice, video, and screen context simultaneously with <300 ms freshness. Candidates who reference these capabilities by name demonstrate they've used the product.
- Prepare for the case format. ** Lucas's case structure, presentation, candidate-led analysis, guided follow-up, mirrors consulting-style interviews. Practice structuring a messy problem aloud in 15 minutes.
- Designers: lead with process. The hiring manager's public tips emphasize story over screenshots: problem framing, constraints, trade-offs, and measured outcomes.
What disqualifies you
- Failing the basics check. ** If the recruiter cannot verify the "main boxes", those requirements, you stop at the phone screen.
- Poor phone-screen hygiene. Background noise, no name introduction, no prepared materials. Recruiter training materials treat these as confidence signals.
- Ghosting the AI screen or demanding a human without asking. The process is built around Lucas; refusing to engage without prior arrangement reads as inflexibility.
- Generic portfolio or résumé. With a small team hiring in bursts (two designers at once, per the LinkedIn post), pattern-matching against the JD is literal. A résumé that doesn't surface the keywords the auto-generated criteria expect won't rank.
The process is fast by design; Tavus's own product pitch is "AI recruiter screening speeds hiring and removes bias with fair, data-driven evaluations." Expect two to three weeks from application to offer if you clear each gate. Ask the recruiter at the phone screen for the exact sequence for your role; the number of rounds varies by function.
Where the work happens
Tavus's headquarters occupies a San Francisco office that functions as the company's primary collaboration hub and the physical anchor for its residency program. The space hosts the core product, design, and go-to-market teams whose roles dominate the hiring mix, along with the lean engineering group building the Phoenix, Sparrow, and Raven model families. Job postings on the Zero G Talent board consistently list San Francisco as the base for salaried positions, including Senior Software Engineer (Infrastructure), Forward Deployed Engineer, Marketer, Customer Engineer, Business Development Representative, and Vibe Growth Marketer, while also offering U.S. remote eligibility for several. The HQ is where the 10-day Tavus Residency runs each summer; the 2026 cohort (July 27–August 5) houses ten builders on-site, covers meals, grants platform access, and culminates in a public Demo Day, turning the office into a compressed product incubator for face-to-face AI prototypes.
A distributed workforce accounts for a meaningful share of headcount. UnifyGTM data identifies an "Other" category of 21 employees, larger than any single named satellite, describing it as a distributed workforce that likely supports remote operations. That figure aligns with the remote-eligible designations on current postings and suggests the company treats remote not as an exception but as a standing operational mode for roles that don't require daily access to the SF lab environment. Revelio Labs puts total headcount at 95 as of March 2026, up 41 year-over-year, so the remote contingent represents roughly one-fifth of the company even before counting any SF-based employees who work hybrid.
Dallas, TX appears in the geographic data with a single employee. The listing gives no indication of a dedicated office; the solitary headcount suggests either a field-based commercial hire or a remote worker who happens to reside there. No current job postings specify Dallas as a location.
Livermore, CA likewise shows one employee. Given its proximity to the Bay Area, this may reflect a commuter arrangement rather than a second facility. No postings list Livermore.
New York, NY rounds out the named locations with one employee. The remainder of the nine locations referenced in aggregate data are unbroken out and carry no public-facing presence.
Who lasts and who leaves
The reviews paint a picture of a company that selects for people who can operate inside a high-velocity, mission-driven environment where the roadmap shifts and the metrics don't always settle. Glassdoor contributors describe "passionate individuals who are deeply motivated by a big, inspiring idea", the kind of language that appears when a founding team (Hassaan Raza, CEO, and Quinn Favret, COO) has articulated a vision sharp enough to attract believers but not yet stable enough to codify. That tension shows up repeatedly: "execution felt unstable," "direction changed frequently," and "success metrics weren't consistent" sit alongside "extremely professional colleagues" and a team that "seem to actually care about their employees work/life." The people who stay are the ones who treat ambiguity as a design constraint rather than a bug.
A high-performance culture, as multiple Indeed reviewers put it, means the pace is real. The commission structure on the sales side is called "VERY good compared to the industry standard," which suggests the company compensates for volatility with upside. Remote flexibility is noted as a genuine benefit, not a perk on paper, and the Mission District HQ anchors a team that still clusters physically when it matters.
What filters people out? One Glassdoor review states flatly: "the culture rewarded urgency over results." That sentence, more than any other, defines the selection pressure. Tavus hires people who can distinguish motion from progress, or who can learn the difference fast. The engineering roles posted to the board (Senior Software Engineer, Infrastructure at $160k–$250k; Forward Deployed Engineer at $140k–$200k) demand technical depth but also the ability to ship against moving targets. The marketing and business development roles, Brand Marketer at $120k–$150k, Business Development Representative, Vibe Growth Marketer, signal a go-to-market motion that values narrative craft and hustle in equal measure. Customer Engineer at $30–$40/hour suggests a support layer that's technical enough to debug in real time.
The through-line across functions: Tavus retains people who find energy in the gap between research and product, who can translate a "new standard for AI-human interaction" (the phrase from the company's own portrait video) into something a customer can buy today, and who don't need a stable process to produce good work. The ones who leave, or say they wouldn't return, tend to be the ones who expected the process to arrive before the outcome.
The kicker: This summer, ten builders will occupy the Mission District office for the Tavus Residency, hacking face-to-face AI prototypes that ship by Demo Day. Lucas will screen the next batch of applicants. The lean engineering team will ship the next model release. The hiring map, heavy on that combination, light on research, will look much the same. The strategy is visible in the roles. The only question is whether the next hire treats ambiguity as a design constraint or a bug.
Working in AI? Zero G Talent tracks the openings: see every open Tavus role, browse AI jobs, the companies hiring, and the people building the field.