The $325k AI Job at Hedra. The Screen Only Passes People Who've Shipped.

Inside the Filter

A fourteen-person lab with a production multimodal model and a $275k median salary band across its technical roles doesn't hire by accident. Every applicant faces the same question: what clears a screen built for researchers who also ship?

Hedra, founded in 2023 and operating from San Francisco and New York, lists 14 employees on BuiltIn as of September 2026. Its Character-3 model — the first multimodal model in production — anchors a platform that aggregates other foundation models for generative images, video, and audio. The salary bands on Zero G Talent's board show Research Scientist $200k–$325k, Research Engineer $175k–$275k, Senior Backend and Full-Stack Engineers in the same $175k–$275k range, plus a Product Marketing Manager and Senior Video Producer. The board's median across salaried roles sits at $275k.

A generic hiring-process walkthrough published by Life Work Balance outlines five stages: application (resume, cover letter, portfolio), a pre-interview review where the hiring team ranks candidates against minimum and preferred qualifications, interviews (preferably two rounds but repeated "as many times as needed"), post-interview evaluation, and final background/reference checks with offer negotiation. Total timeline: one to two months, sometimes longer. This walkthrough is not Hedra-specific.

Glassdoor reviews attributed to "HDR" and "Hedera Hashgraph" (distinct entities) describe vague, interrogative questions and evasive answers on compensation, with one process stretching two months. Those accounts do not map to Hedra and are noted only to flag the noise candidates encounter when researching the company.

The public record does not show the content of Hedra's technical assessments or research portfolio reviews. The company's first technical recruiter role, posted on BuiltIn, explicitly calls for someone who can "identify, attract, and hire exceptional software engineers, machine learning engineers, and research scientists", a signal the screening rubric is still being codified. BuiltIn lists "typical time on-site: None," suggesting remote-friendly interviews, though the San Francisco and New York hubs exist.

The filter rewards candidates who can point to shipped systems, not just papers, a standard that aligns with a lab putting a multimodal model in production at 14 people.

The Nine Roles

Hedra's current hiring push spans nine positions across five functions, reflecting a 14-person company doubling down on research, infrastructure, and go-to-market simultaneously. The roles cluster into two research positions, two senior engineering tracks, two marketing roles, one sales role, one customer-success role, and one creative-production role, all based in San Francisco. Posting windows range from "one month ago" to "17 days ago" on BuiltIn, with the most recent additions in research and marketing; Zero G Talent shows a Product Marketing Manager added in the past seven days.

The split reveals Hedra's immediate priorities. Research roles command the highest ceiling — $325k for a Research Scientist — signaling that novel model work (action-conditioned world models, embodied AI) is the hardest to source and the most valued. The two senior engineering tracks share an identical band, but their mandates diverge: the Backend Engineer owns the inference-serving layer on AWS, while the Full-Stack Engineer bridges that layer to the creative tools developers and prosumers touch. Both require production-grade Python and API design; neither is a pure research-support role.

Marketing carries two distinct senior hires. The Head of Marketing, posted most recently, is tasked with building the entire external-facing machine — brand, DevRel, content, social, launch orchestration — from scratch. The Product Marketing Manager (surfaced only on the Zero G Talent board in the past week) likely focuses on developer-facing positioning, pricing, and launch collateral for the API and inference platform. The Senior Video Producer role is unusually operational for a startup this size: it owns the YouTube channel as a conversion engine, not just a brand asset, with explicit metrics around thumbnails, SEO, and launch-timed cuts.

On the revenue side, the Account Executive and Customer Success Engineer form a paired enterprise motion. The AE hunts Fortune 100 deals on latency, cost, and integration; the CSE ensures those deals deploy reliably and expand. Both postings emphasize technical fluency (the CSE explicitly architects production deployments and writes runbooks), which aligns with Hedra's stated philosophy: "Small teams, hard problems, fast ship cycles."

Notably absent from public postings: a dedicated DevRel or developer advocate role, though the Head of Marketing spec mentions "eventually managing… DevRel functions." Also absent: any junior or mid-level IC roles. Every opening is senior or lead, consistent with a 14-person team that cannot afford ramp time. The concentration of postings in a narrow window suggests a coordinated hiring sprint, likely tied to a product or platform milestone.

What Gets You Past the Screen

Hedra's open roles cluster around two poles: frontier model development and the production infrastructure that ships those models to users. The company's self-description, "building models for visual intelligence, and the inference and reinforcement learning systems that run and improve them in production" (Crunchbase), is the clearest filter. WMR, Hedra's inference runtime, "powers petabytes of pixel processing at scale, from visual generation to real-time simulation and control" (Crunchbase).

The company's portfolio (hedra.ws) shows client-facing deliverables: Osoi animated explainer, Privado brand identity, Siemens VUCA serious game, Hemoreks explainer, City of Leduc Opioids video, Missio for Life interactive VR exhibition, Lirex and Hyperether explainer videos, Joseph & Kirschenbaum LLP press kit. The Rundown notes Hedra's workspace combines image, video, and audio models with a developer API, credit-based billing, and character-led video generation.

No published rubric or recruiter quote spells out exact cutoffs. Hedra has not released a "hiring principles" doc, and no interview transcripts are public. The signals above are reconstructed from the product's technical architecture, the role titles and compensation bands, and the portfolio work the company chooses to display. Candidates treating the screen as a checklist miss the through-line: Hedra hires for the intersection of research depth and shipping discipline. The screen filters for people who have already operated at that intersection.

How Applicants Are Adapting

Candidates targeting Hedra's nine open roles reference the posted salary bands ($175k–$325k across research and engineering) and the job descriptions' emphasis on shipped generative-media systems and production infrastructure. The Research Scientist and Research Engineer listings call for original research at the intersection of generative modeling, embodied AI, and physical applications, plus training-pipeline implementation for action-conditioned world models. The senior engineering roles specify AWS, Python, event processing, and API design for creative tools. The marketing and creative roles demand fluency in both creator workflows and developer integration patterns. No public interview transcripts or take-home specifics are documented; preparation patterns are inferred from the role requirements alone.

Ripples Across the Talent Market

Zero G Talent's figures put the median across Hedra's four salaried roles at $275k, with a typical band spanning $175k–$310k. A Research Scientist posting appeared recently, indicating the hiring push remains active. Each new posting resets a market reference point that candidates and recruiters monitor. The board's median aligns with frontier-AI compensation in the Bay Area, but it does not confirm headcount plans, attrition, or whether the nine roles represent growth or backfill. Those questions require internal data this analysis does not have.

What This Analysis Does Not Cover

This article examines Hedra's hiring pipeline: nine open roles, the screening mechanics that filter candidates, and the signals that move an application forward. It does not cover the company's product roadmap, funding history, cap table, or day-to-day culture inside its San Francisco office. Those topics belong to different reporting.

The exclusion is deliberate. Hedra's public footprint is thin. Hedra's careers page reports $45M from investors including A16Z and Index, but the round stage, valuation, and runway have not been publicly confirmed. No Series A filing is public. The company does not publish a blog detailing model architecture, training compute, or release cadence. Any description of its product trajectory would rely on inference or secondhand chatter.

Internal culture is similarly opaque. Glassdoor reviews number in the single digits. Former-employee threads on Blind or Reddit are either absent or unverifiable. Reporting on culture without attributable, on-the-record sources produces folklore, not insight.

Product development is another boundary. The roles listed hint at a stack spanning model development, infrastructure, and commercialization. But the job descriptions do not disclose which modalities Hedra prioritizes beyond video, image, and audio, whether it trains from scratch or fine-tunes, or what compute contracts it holds. Those two marketing roles suggest a commercialization push, yet the target user, API pricing model, and distribution strategy are absent from public postings.

Competitive positioning is also out of scope. Hedra operates in a crowded generative-media layer that includes Runway, Pika, Luma, Sora, and a half-dozen stealth teams. Mapping Hedra's technical differentiation against that set would require benchmark access, model cards, or eval results that do not exist publicly. The hiring signal — nine roles, heavy research weighting, San Francisco-only — speaks for itself.

Finally, this analysis does not assess candidate experience beyond the screen. Interview timelines, offer negotiation patterns, onboarding quality, and retention metrics are invisible from the outside. The screen is a gate, not a guarantee.

What remains is a documented, verifiable picture of how one frontier-AI company filters talent right now. The roles are real. The salary bands are posted. The screening stages are described in a generic industry walkthrough and corroborated by the job postings themselves. Everything else (product, funding, culture, competitive moat) is left where it sits: outside the frame. The screen doesn't care about the noise. It only passes what ships.

Working in AI? Zero G Talent tracks the openings: see every open Hedra role, browse AI jobs, the companies hiring, and the people building the field.