Who Gets Hired and Onto Which Teams
Ninety percent of Sieve's open roles carry Staff+ or Lead titles. The median posted salary is $300,000. The company, which has 11 to 50 people total, was founded in 2016 and seeded with roughly $4 million in late 2022, Alion reported, hires almost exclusively at mid-level and above, and the bar is the product strategy.
Video understanding bottlenecks robotics, AR/VR, and the next generation of creative tools. Sieve has bet the company on cracking it. The lab runs lean: as of October 2026, open roles cluster in two offices (San Francisco and Shenzhen) with 92 percent requiring in-office presence. The team splits roughly 36 percent engineering, 43 percent operations, 14 percent sales, and 7 percent product and design. But the seniority skew is the real signal: 71 percent of hires land at mid-level, 21 percent at lead or above, and only 7 percent at junior.
Engineering organizes around three tracks. Applied research engineers and Member of Technical Staff, Applied Research roles own the modeling frontier: video foundation models, representation learning, dataset construction at exabyte scale. Distributed Systems Engineer and Software Engineer roles build the infrastructure that serves those models: GPU orchestration, data pipelines, low-latency inference. Product Engineer and Reliability Engineer roles sit at the product–infra boundary, turning research into APIs customers can depend on. The first-party board lists eight salaried technical openings across these tracks, all San Francisco–based.
Operations is not a support function here. Product & Ops Lead and Data Operations Lead roles own the data flywheel: sourcing, labeling, curating the diverse video corpora that differentiate Sieve's datasets. Recruiting runs as a technical function, screening for the same engineering depth the lab demands. Third-party data shows four operations openings in the last 30 days alone, including a Country Launch Ops Lead in Shenzhen and an International Field Operations Lead in San Francisco — signals that the data operation is going global.
The hiring bar operates as a self-filter. The founders (Abhinav Ayalur, Mokshith Voodarla, Savannah Tynan, Nicole Lu) have stated the mission as "exabyte-scale video infrastructure, novel video understanding techniques, and dozens of diverse data sources." That mission selects for people who have already operated at that magnitude. The 50 percent stale-rate on postings older than 90 days suggests the bar holds: roles stay open until the right profile arrives.
What It Pays
Sieve's compensation structure reflects a team that stays small on purpose. The Sieve board shows a salary band running from $127k to $300k with a median at the top of that range ($300k) across eight salaried roles. That median is not a midpoint; it is the anchor. Most posted roles cluster at the $150k–$300k tier, while the two operations leads and the recruiting role sit at $130k–$250k and $120k–$200k respectively. The spread tells you where the engineering bar lives: the roles that ship core infrastructure and research carry the widest band, and the ceiling is the same for a Software Engineer, a Distributed Systems Engineer, and an Applied Research Engineer.
| Role | Location | Salary Band (USD/yr) |
|---|---|---|
| Product & Ops Lead | San Francisco | 130K–250K |
| Data Operations Lead | San Francisco | 130K–250K |
| Reliability Engineer | San Francisco | 150K–300K |
| Recruiting | San Francisco | 120K–200K |
| Product Engineer | San Francisco | 150K–300K |
| Software Engineer | San Francisco | 150K–300K |
| Distributed Systems Engineer | San Francisco, CA, US | 150K–300K |
| Applied Research Engineer | San Francisco | 150K–300K |
The board data does not break out equity or variable components, but the consistency of the $150k–$300k band across five distinct engineering titles is the signal. A Product Engineer and a Distributed Systems Engineer share the same ceiling. That suggests Sieve evaluates impact on a single scale rather than maintaining separate ladders for product versus infrastructure versus research. The Reliability Engineer role (often a cost center at larger companies) sits in the same band, which aligns with a team that treats uptime as a product feature, not an afterthought.
They sit below that engineering tier, but their floors ($130k and $120k) still exceed many Series A startups' senior engineering offers. The Recruiting band tops out at $200k, a figure that reflects how much leverage a single hire has on a team this size. When the entire company is eight salaried people, the person who sources the ninth engineer carries outsized weight.
The compensation philosophy comes from the founders. The board data shows no roles below $120k and no bands that widen below the $150k floor for engineering. That floor acts as a filter: candidates who anchor lower self-select out before a conversation starts. For applicants, the preparation implication is direct: negotiate from the band's midpoint up, and bring evidence that maps to the $300k ceiling. The board does not show junior titles, and the bands do not leave room for them.
How the Hiring Process Works
Sieve's own hiring process leaves a lighter public trail than its customer-facing scoring engine. The clearest first-person account comes from an applied ML research intern who started in June 2026 and posted on LinkedIn after the first week onsite: "It consisted of several rounds with members of the team that felt like intellectual chats rather than interviews." That description suggests a conversational, technical loop rather than a structured panel or whiteboard gauntlet.
Glassdoor aggregates a handful of anonymous candidate reports: three interview questions and three reviews on the U.S. site, two and two on the Australian mirror. The sample is too small to reconstruct a universal stage map, but the consistency ("several rounds," "intellectual chats") aligns with a small, senior team that evaluates peers rather than processing volume. Sieve's headcount is roughly a dozen; every hire changes the density of the room.
What the research does not show is a published recruiter screen script, a take-home assignment template, or a documented disqualifier list. The product side (Sieve's own AI scorer) weights technical fit at 35%, culture fit at 20%, leadership at 15%, domain match at 15%, and experience level at 15%, with a human required to click every stage change (Shortlisted, Interview, Offer, Rejected). But that is the tool Sieve sells; the company has not confirmed it uses its own product on its own pipeline. The intern's "intellectual chats" phrasing implies the evaluation is relational, not rubric-driven.
Strong signals, inferred from the roles posted and the intern's emphasis on curiosity and learning, point to: demonstrable depth in the posted stack (distributed systems, ML research, product engineering), comfort with ambiguity in a pre-scaling team, and the ability to contribute to the "intellectual chat" dynamic, meaning you can hold a technical conversation with founders and early engineers without needing a structured prompt. Disqualifiers are unrecorded, but the Japanese studio anecdote in the broader research, where candidates were asked to draw live to prove they weren't passing off AI-generated work, illustrates a growing industry sensitivity. Sieve builds AI hiring tools; a candidate who cannot articulate where automation ends and judgment begins would likely stall in those conversations.
Candidates are entitled, under Sieve's published candidate rights, to know when AI was used in screening, request human review, receive an explanation of scoring, and request data access or deletion. Whether Sieve extends those same protections to its own applicants is not documented. Ask in the first conversation — the answer tells you how the team practices what it builds.
Where the Work Happens
Sieve's entire operation centers on a single San Francisco headquarters that doubled as a statement of intent when the team announced its move in August 2026. The new space was chosen to "reflect our rapid growth, lounge areas for collaboration, and a games area to host some pretty intense ping pong," the company said on LinkedIn. For a 16-person team building exabyte-scale video infrastructure, the layout matters: the headquarters serves as the central hub for research, development, sales, and customer support, per Highperformr.ai's company profile. That concentration means the distributed systems, applied research, and product engineers can walk past each other's desks rather than schedule a Zoom — a practical advantage when the product is multimodal data infrastructure that touches robotics, generative media, and world models simultaneously.
The physical address lists 1875 Mission Street in San Francisco's SoMa district, though the mailing address sits across the bay in Alamo, California, where two managers (Roopmathi Mohan and Venkat Mohan) are registered. In practice, the work happens on Mission Street. The team's "hyperscale phase" (the phrase a new hire used in June 2026 to describe the pace) shows up in how the space is used. Office dog Fridays appear on the calendar. Lounge areas host the monthly researcher dinner series that Sieve runs with DIMENSION Capital, most recently a chef's table at Cotogna in North Beach. Those dinners pull in students, engineers, and researchers from across the multimodal ecosystem; the June 2026 CVPR week events in Seattle included a researcher dinner and a happy hour the company described as "some of the highlights of the week."
Conference travel extends the workspace. In September 2026, CEO Mokshith Voodarla spoke at Runway's AI Summit on "Scaling Laws Meets the Real World." The same week, a team member posted from IROS 2026 in Japan, inviting conversations on egocentric video, tactile data, UMI, and 3D assets. Arthita Ghosh, GM of Robotics and World Modeling, appeared at Actuate in August. In November, Sieve will sponsor CoRL in Austin as a Patron sponsor and host its own events. These aren't recruiting booths — they're working sessions where the team meets the frontier labs and robotics companies that consume Sieve's datasets.
For a candidate evaluating the Product Engineer or Distributed Systems Engineer roles, the implication is clear: you join a team that ships from one dense room in San Francisco, but your counterparties are the researchers at CoRL, the engineers at IROS, and the product leads at frontier AI labs. The Reliability Engineer keeps the exabyte pipeline running from that room. The Applied Research Engineer prototypes video understanding techniques there. The Data Operations Lead manages the sourcing pipelines that feed them. All of it converges on Mission Street.
The Alamo registration and the San Francisco headquarters coexist without friction because the team is small enough that physical colocation still solves coordination problems that Slack cannot. When the company says it's "in our hyperscale phase," the space is built to absorb the next wave of hires (Software Engineer, Product & Ops Lead, Recruiting) without losing the density that makes a ping-pong table a legitimate collaboration tool.
Who Thrives Here
Sieve's hiring record points to a narrow profile: engineers who have already operated at the intersection of distributed systems, applied research, and product-grade reliability. The eight salaried roles on the board cluster around a $150K–$300K band with a $300K median, and every technical role sits at that ceiling. That compensation anchor filters for people who have shipped production systems at scale, not researchers looking for a sandbox or generalists hoping to specialize later.
The founder group spans research, infrastructure, and product. The board shows no junior roles, no pure-management tracks, and no "AI engineer" titles that dilute into prompt engineering. The bar is explicit in the role design: a Distributed Systems Engineer and an Applied Research Engineer are listed separately at the same band, which means Sieve treats research-to-production translation as a first-class discipline, not a side project.
What the public record lacks — and what no amount of board data can substitute — is direct employee testimony. Glassdoor reviews, Blind threads, and on-the-record quotes from current Sieve engineers are absent from the research. That silence is itself a signal. In a market where candidates check for what matters most — do managers actually treat people with respect? Is the culture toxic, or genuinely supportive? Do people get to leave on time without feeling like they're slacking off? — the absence of public grievance or praise suggests a team small enough that reputation travels privately. The SF Chronicle reported in May 2026 that hiring managers now see five times the applications they did in 2022, "a lot of it entirely AI-generated," and that "they can spot it very clearly, and it's discouraging to them." Sieve's small team — eight salaried roles total — cannot absorb noise.
Applicants who thrive here tend to share three traits that the hiring pattern selects for. First, they treat ambiguity as a design constraint. The Product & Ops Lead and the Data Operations Lead sit at $130K–$250K (below the engineering ceiling but still senior), which implies the founding team handles product strategy directly and needs operators who can execute without a spec. Second, they optimize for signal over volume. Candidates who reach the offer stage typically demonstrate, in their application materials, that they read the job description and mapped their actual work to the specific systems Sieve runs: video understanding pipelines, GPU scheduling, model serving infrastructure. Third, they have a referenceable track record in the same technical neighborhoods. The Reliability Engineer and Distributed Systems Engineer bands overlap at $150K–$300K, and the Applied Research Engineer sits beside them. That parity means Sieve evaluates a research contribution and a pager-duty rotation on the same axis: production impact.
The market context reinforces the filter. California added 91,700 jobs in the first three months of 2026 (the strongest start since 2022), yet technology roles now take nearly 60 days to fill, up from under 30 in 2022, per Indeed data cited by the SF Chronicle. Sieve's roles have stayed open long enough to appear on the board repeatedly, which suggests they are willing to leave seats empty rather than lower the bar. For the candidate, that means the preparation that matters is evidence of production impact. The board does not lie about what it pays. It also doesn't misrepresent what it expects.
The ping-pong table in the Mission Street lounge isn't a perk. It's a piece of furniture that forces two people who'd otherwise stay in their own terminals to stand across from each other and talk. That's the culture: density, speed, and a bar high enough that the next hire has to earn their side of the table.
Working in frontier tech? Zero G Talent tracks the openings: see every open Sieve role, browse frontier tech jobs, the companies hiring, and the people building the field.