Alexis Gauba

Alexis Gauba is a co-founder and chief operating officer of Raindrop, which monitors AI agents for silent failures and connects alerts to conversations and traces engineers can investigate. She founded the company alongside Zubin Koticha and Ben Hylak. Her public writing emphasizes preventing consequential failures and preserving context between agents that investigate problems and agents that fix them.[2][3][4][5]

In a June 17, 2026 post, Gauba argued that evaluation should prioritize the moments when an agent confidently does the wrong thing. Her prescription was to “stop being a benchmark-maxxer and start being a floor-raiser.” By floor-raising, she meant identifying catastrophic failures and preventing their recurrence; she favored an agent acknowledging uncertainty over hallucinating with confidence.[4]

Her customer writing makes the gap between model behavior and user experience concrete. Raindrop credits Gauba with its November 30, 2025 Spiral case study: the writing agent called a tool to create a document, but a rendering problem prevented users from seeing it. The model believed it had succeeded, while frustrated users prompted repeated tool calls. Spiral’s general manager Danny Aziz explained that other logs would have missed the problem because “from the model's perspective, it was totally fine.” The case study reports that Spiral fixed the bug the day it was reported and that it had affected around one-third of users during its V3 launch week.[3][8]

On May 6, 2026, Gauba explained Raindrop’s collective decision to expose its debugging agent, Triage, as a tool through the Model Context Protocol (MCP). Calling the specialized agent lets coding agents use its tuned debugging harness and accumulated knowledge about which issues matter to a customer. In the launch announcement, authored by CTO Ben Hylak two days earlier, the team explained that this arrangement gave it control over Triage’s model and harness while supporting different coding agents. Triage’s underlying tools also remained available through MCP.[5][6]

Gauba illustrated the context handoff with a Slack workflow: post a bug, ask Raindrop to investigate, then ask Devin to fix it. Devin can call Triage to continue the existing investigation, carrying forward the customer’s priorities and earlier discussion. She called this approach the “Multi-Agent Covenant.”[5][6]

On September 17, 2026, Gauba, Koticha, and Hylak jointly announced Raindrop Simulations, extending the company’s work from detecting production failures to testing proposed agent changes. They described replaying production traffic and existing tests against a changed harness, then applying anomaly detection to identify unexpected behavior. The technical challenge, they explained, was simulating the agent’s surrounding environment rather than simply replaying cached tool responses—particularly when a change introduces a new tool. The announcement broadened early access and anticipated general availability over the following month.[7]

Gauba also hosted the AI Design Engineering track at the 2025 AI Engineer World’s Fair, receiving credit among the track hosts in Latent.Space’s June 13 recap.[1]

Before Raindrop, Gauba co-founded Opyn, a decentralized-finance options platform. That earlier work provides context for her experience founding technical companies. She also co-founded she256 alongside Sara Reynolds and Medha Kothari. The organization works to increase diversity and lower barriers to entering blockchain through access to industry professionals and education for underrepresented people.[2][9]