Staff / Senior Staff Engineer, AI Products

We’re a team of ex-Google engineers who built some of the largest defensive platforms on the planet — Safe Browsing and reCAPTCHA. Now, we’re striking out on our own to tackle an even bigger challenge: stopping the new wave of adversarial AI attacks already hitting organizations today. We're going after a $5B+ market, ripe for disruption. Traditional detection methods are too slow to keep up. Adversaries are using AI to craft customized, high-evasion attacks — and old-school rules-based systems don’t stand a chance. We’re looking for a Staff or Senior Staff engineer to help build the next generation of security products at Aegis. This is a highly technical role with a path to significant leadership. You won’t walk in and inherit an organization because of your title. You’ll start by owning hard problems, shipping, setting technical direction, and making the engineers around you better. As you prove that you can lead the product and the people building it, the scope can grow quickly. If you’re a Staff or Senior Staff engineer who wants to build a team—or an engineering manager who wants to get closer to the technology before taking on much larger leadership scope—this role is designed for that trajectory. NY/SF in-person preferred, Open to Remote Your Mission Build security systems that can make good decisions from incomplete and noisy information. Modern security products rarely get a perfect signal. They have to combine behavioral patterns, historical context, content, infrastructure signals, and other telemetry to determine whether something is actually risky—and decide what to do about it. You’ll work on problems like: Turning large volumes of noisy security data into high-confidence detections. Influencing/Participating in model creation to build the best product Understanding what “normal” looks like when it is different for every user and organization. Moving new detection ideas from prototype to reliable production systems. Measuring whether the systems we build are actually correct. The systems you build will operate in environments where false positives have consequences and latency, precision, reliability, and customer trust all matter. You’ll work across detection systems, distributed infrastructure, product surfaces, and AI inference—and you’ll be expected to understand the full path from raw data to a customer-visible decision. Why This Is Hard There is rarely a clean signal. Real-world security problems are ambiguous. The same behavior can be harmless in one context and dangerous in another. The interesting engineering work is figuring out which signals matter and combining them into decisions you can defend. “Normal” constantly changes. Every organization behaves differently. Users change jobs, locations, devices, tools, and workflows. New customers arrive without much historical data. The systems need to learn without becoming brittle. Precision is part of the product. A security system that catches everything but constantly cries wolf isn’t useful. You’ll build evaluation systems, labeled datasets, replay infrastructure, and production feedback loops alongside the detection itself. AI has to work in production. It’s one thing to demonstrate an impressive model offline. It’s another to make AI part of a production system with real latency, reliability, privacy, and cost constraints. Detection is only part of the problem. The most useful security systems help customers act. Building automated or assisted responses that customers trust requires a much higher engineering bar than producing another finding in a dashboard. Some of the product still needs to be invented. You’ll have meaningful influence over what we build, which problems we prioritize, and how early technical ideas become products. There is no giant organization between you and the problem. If you like ambiguous technical problems, fast iteration, and being responsible for whether the thing actually works in production, you’ll have a lot of fun here.