Arena has six open roles. All six are senior engineering positions. All six sit in the Bay Area. All six pay between $150,000 and $350,000 in total cash compensation, Zero G Talent's job board reports.
This guide maps what that board reveals — who Arena hires, what they earn, how the interview loop works, where the work happens, and what the public record shows (and doesn't) about who lasts.
Who gets hired
Every open position carries a "Senior" or "Staff" prefix. That concentration reflects a product strategy leaning hard on infrastructure, machine-learning systems, and trust-and-safety tooling rather than a broad functional spread. The six roles map to five engineering teams, detailed in the compensation table below.
Notably absent: design, product management, data science, sales, marketing, operations. Either those functions hire through other channels, contract out, or don't exist at meaningful scale yet. The board captures a snapshot — six roles, all engineering, all senior, all local — but doesn't reveal whether Arena hires into leadership, supports internal mobility, or runs university recruiting. Those questions belong to the hiring process itself.
The pay picture
| Role | Level | Base Range (USD/year) |
|---|---|---|
| Staff Software Engineer, Product | Staff | $200,000 – $350,000 |
| Senior Software Engineer, Product | Senior | $150,000 – $300,000 |
| Senior Software Engineer, Data Infrastructure | Senior | $150,000 – $350,000 |
| Senior Software Engineer, ML Infrastructure | Senior | $150,000 – $350,000 |
| Software Engineer, Product Security | Senior | $150,000 – $350,000 |
| Software Engineer, Anti-Abuse | Mid/Senior | $180,000 – $300,000 |
All figures reflect total cash compensation as posted; equity and benefits are not broken out. The $325,000 median implies the distribution skews toward the upper half of each band, consistent with a company hiring experienced engineers into high-leverage roles rather than filling junior slots.
Bay Area pricing drives the ceiling. A $350,000 top end aligns with competitive staff-level packages at nearby AI labs and platform companies. The $150,000 floor exceeds typical senior-engineer entry points in lower-cost metros. Treat the posted ranges as negotiable bands, not fixed grids; the 200-point spread between floor and ceiling on most roles leaves meaningful room for level calibration, competing offers, and domain scarcity: ML infrastructure and anti-abuse expertise command premiums.
No public filing or company blog breaks out Arena's equity structure, refresh cadence, or sign-on norms. The board captures only the annual cash figure advertisers chose to disclose. For a complete picture, ask recruiters to specify the RSU grant schedule, vesting cliff, and whether the upper-band cash number assumes a target bonus or is pure base.
Inside the interview loop
Arena's public-facing hiring documentation is thin. The company's own blog points candidates to arena.ai/jobs with a single line, "Join the team," and lists the same six Bay Area engineering positions. Beyond that, Arena has not published a recruiting playbook, a careers FAQ, or a breakdown of interview stages.
What exists in the public record is a generic five-step framework described in a 2023 YouTube walkthrough: apply → pre-interview screening → interview(s) → post-interview evaluation → offer and negotiation. That video notes the full cycle typically runs "about a month to two months sometimes more." Screening usually filters first for minimum qualifications, then for preferred qualifications, with candidates who meet preferred criteria ranked highest. The same source says two interviews are often enough to assess skills. None of this material is Arena-specific; it reflects broad tech recruiting norms.
From the role titles and levels on the job board, the pipeline likely includes a technical screen (coding or system-design) followed by deeper-dive rounds with the hiring team. Staff-level postings suggest at least one round evaluating architectural judgment and cross-functional collaboration. Security and anti-abuse roles imply domain-specific assessments (threat modeling, abuse-pattern analysis) may be part of the loop for those tracks. But without a published process or on-the-record comments from Arena recruiters or hiring managers, these are deductions from role requirements, not documented steps.
What distinguishes successful applications at most AI-infrastructure companies, and what Arena's role descriptions signal, is demonstrable production experience with the stack each team owns: distributed data pipelines for Data Infrastructure; large-scale model serving and training infrastructure for ML Infrastructure; product-facing API design and experimentation frameworks for Product Engineering; adversarial thinking for security and anti-abuse. The salary bands align with senior-and-above IC expectations, so candidates who clear the first screen typically show ownership of end-to-end systems, not just component-level contributions.
What predicts tenure after hire isn't in the public record — only the entry filters are visible.
Where the work happens
Every role on Arena's board carries the same location tag: Bay Area. The company's roots are in the East Bay. Arena grew out of a PhD research project at UC Berkeley, where the founding team built Chatbot Arena as an academic evaluation platform. The rebrand announcement in January 2026 noted the platform had reached more than 5 million monthly users across 150 countries and 60 million conversations per month, but the blog post did not disclose a physical headquarters address. Arena's own site and public materials have not published an office lease, floor plan, or return-to-office policy.
The broader San Francisco AI office market provides context. Since 2020, AI companies have leased nearly 7 million square feet in the city (roughly one square foot in twelve of total office stock) according to JLL data cited by The San Francisco Standard. The densest clusters sit in Yerba Buena, SoMa, South Beach, and the Financial District, spilling into Dogpatch, Mission Bay, and Potrero Hill. Mission Bay, where OpenAI now occupies more than 1 million square feet around the Chase Center, carries a 9.1% vacancy rate and Class A rents around $125 per square foot. Showplace Square and the Design District run at less than half that rate; brokers say lower-priced markets with similar freeway access are absorbing demand.
Arena has not been named in public lease filings or the Standard's mapped footprint. JLL's dataset obscured company names for confidentiality and excluded small companies working from incubators or flex spaces. Arena's headcount (implied by six salaried roles on the board) would place it in that smaller tier. The research notes that many AI startups operate "in plain sight" with obscured signage and addresses, citing fears of corporate espionage, recruiting wars, and staff security.
Y Combinator held its 2026 Startup School at the Chase Center in Mission Bay, a 6,000-attendee event in the heart of the AI office cluster. The accelerator's travel support for non-local founders underscores that the Bay Area remains the default gravity well for early-stage AI companies. But attendance at a YC event is not a lease.
For a candidate, the practical takeaway is straightforward: plan on a Bay Area presence. Whether that means a daily commute to a SoMa loft, a hybrid schedule from a Mission Bay desk, or a quarterly on-site from elsewhere in the region, the job postings treat the Bay Area as the hiring geography. Remote-only applicants should clarify expectations early; the board does not list a single role tagged "remote."
Six roles. Six senior engineers. One Bay Area address. The board is the only contract Arena has published — and the only one candidates can verify.
Working in frontier tech? Zero G Talent tracks the openings: see every open Arena role, browse frontier tech jobs, the companies hiring, and the people building the field.