About CivAI

CivAI (https://civai.org) is a nonprofit that disseminates knowledge about AI capabilities and dangers in a unique way. Instead of research papers and declarative statements, we produce interactive software experiences that provide a deep, intuitive sense of where AI is and where it’s going. We present these demos to targeted audiences—policymakers, interest groups, conferences—and make scalable versions available to the public.

Since starting in October 2023, we have built demos of AI-generated phishing, election interference, one-shot deepfakes, and more. We have presented these to dozens of groups, including federal and state congressional offices, elections officials, and trade associations. Press coverage of our work, including mentions in the New York Times and Washington Post, can be found here: https://civai.org/press

CivAI is aiming for a world where people deeply grasp AI capabilities, and make better decisions because of it. This is a world that holds AI developers responsible for the externalities they are producing, while also finding use cases that transform lives for the better. CivAI’s work brings a new kind of evidence to the discourse—simple, intuitive, and incontrovertible because users can interact with it firsthand.

We'd love to hear from you! If you have any questions about this role, please email us at contact@civai.org.

About the role

As a software engineer, you will be responsible for conceiving of, and building end-to-end, the software demonstrations that are the core of what we do. These will explain both prosaic AI capabilities and more nascent ones that are fast approaching.

At CivAI, we are applying the toolkit we know best, having previously founded a successful software company—developing software products in a startup environment. Consequently, we believe this role is an opportunity for a generalist engineer to do impactful work building user-facing software products outside of a research setting.

In this role, you will:

• Build our demonstrations from end to end, acting as an agentic, creative, and conscientious hacker on a collaborative team.
• Write quality software. That includes considered UI/UX, products that are robust to misuse, and performant code—we want our public demos to be used by thousands of people.
• Come up with new demonstrations that our audiences will find meaningful. See the Representative Projects section for more.
• Use a wide variety of frontier AI systems (commercial LMs, open models, duct-taped research libraries, GPU providers) and tinker with them when necessary.
• Maintain an awareness of AI progress, in order to educate our audiences and inform new demonstrations.
• Iterate quickly on lightweight projects that pack a rhetorical punch, rather than toil for months on a surgical change to an enterprise codebase.
• Communicate understanding about AI through the software and its accompanying text, in the spirit of interactive explainers or data journalism.

You may be a good fit if you:

• Want to help the world understand AI’s implications, so that future AI progress is deliberate and beneficial for humanity.
• Have built complete software projects that you’re proud of, especially full stack web applications. These can be from jobs, side projects, hackathons, or internships.
• Enjoy brainstorming about creative ways to showcase how AI is going to manifest in the world.
• Care about perfecting the entire user experience, from the look and feel of the UI to the speed of the AI inference.
• Have an opinionated aesthetic sense of what makes a good software API or user interface.
• Are willing to roll up your sleeves and try the kitchen sink if something isn’t working. We frequently work with cutting edge research code that is sparsely maintained.
• Practice security mindset, which will mean coming up with exotic ways our products could be misused and guarding against them.
• We are open to varying levels of experience for this role, but strong candidates may have 3+ years in a company setting.

Representative projects

• Our policymaker demos suite, which starts with someone’s LinkedIn and makes instant deepfakes of them using a combination of open models, writes personalized phishing emails with Gmail-like UI, creates fake news Tweets and NYT articles, and more.
• A “talking head” AI avatar realistic enough that it feels like a video call. This would likely require use of novel research code.
• Displays of agentic behavior, persuasion, and large-scale action. For example, the TaskRabbit test carried out by OpenAI/METR. Evals can inspire lots of demos.

Salary and benefits

• The role is full-time and on-site at our office in Berkeley.
• The salary range is $115,000 to $145,000.
• We provide catered lunch and dinner at our office every weekday.
• We are unable to sponsor visas at this time but hope to do so in the future.

Application process

• Initial application (15 min) — Resume and optional project links to assess fit.
• Come up with a new demo (1 hr) — We'll send you a short task to imagine and technically scope a new AI capabilities demonstration.
• Informational discussion (45 min) — We’ll chat about your demo idea, show you what we’ve built, and share more about CivAI.
• Technical phone interview (1 hr) — Building something cool on a video call.
• Remote final round interview (3-4 hrs) — Two coding interviews, and conversations with both founders.

CivAI is an equal opportunity workplace, and we welcome applicants from all backgrounds regardless of race, color, religion, gender identity, sexual orientation, national origin, age, socioeconomic status, disability, veteran status, or any other characteristic protected by law.