Fonzi AI to host agents reliability event

Fonzi AI promoted a New York AI Engineers event for Wednesday, May 20, focused on how engineers are trying to make AI agents more dependable in production. A Luma event page listed the session as a two-hour evening gathering in Kings County, New York, with technical talks and Q&A. The speaker lineup included engineers from Hebbia and Noemica, as well as Harmonic and Instacart. The event description said the program was designed for engineers working on real-world AI systems rather than general-interest product demos. ### What exactly was Fonzi AI hosting? The May 20 listing described the program as the latest edition of “New York AI Engineers Tech Talk,” a recurring meetup focused exclusively on AI engineering. The page said attendees would hear speakers discuss systems they had built, the technical problems they ran into and how they addressed them. The agenda on the Luma page scheduled networking from 6:00 p.m. to 6:20 p.m. (luma.com) EDT, opening remarks at 6:20 p.m., two rounds of 10-minute talks with five-minute Q&A, and speaker meet-and-greet sessions through 8:30 p.m. The same page said Fonzi AI was providing pizza, beer and wine. ### Where did “reliable AI agents with feedback loops” fit into the program? (luma.com) Yesh Chandiramani of Harmonic was listed for a talk titled “Building Reliable AI Agents with Feedback Loops.” The event page said the presentation would cover how Harmonic is building Scout, an AI assistant used in “hard-to-evaluate domains like market mapping and investor search,” and what it takes to design feedback loops that improve output quality. (luma.com) The same event notice paired that session with an Instacart talk titled “Catching Model Failures Before They Ship.” That description said the talk would examine eval pipelines that surface hidden regressions, reduce reliance on slow human labeling and speed up iteration cycles. ### Which Hebbia and Noemica speakers were named? The event listing named David Osemwegie, a member of technical staff at Hebbia, as one of the speakers. (luma.com) His talk, “Inside Enterprise-Grade Coding Agents,” was described as a breakdown of the architecture, security and design behind Hebbia’s background agent and how teams use those systems in production. The same listing named Sebastian Sosa of Noemica for a session titled “Why Most AI Agents Are Built Wrong (and How to Fix Them).” The page described that talk as a “practical teardown” of common mistakes in agent design, along with patterns used by teams at OpenAI, Anthropic and Google. (luma.com) ### Why is Hebbia relevant to a discussion about agent reliability? Hebbia’s own engineering materials show the company has been publishing work on how to measure agent quality. (luma.com) A March 26 post on Hebbia’s blog was titled “Evaluating AI Agents: A Hybrid Deterministic and Rubric-Based Framework,” and described how the company measures agent quality at scale. OpenAI also described Hebbia as running a multi-agent system for finance and legal work. An OpenAI customer story said Hebbia’s Matrix platform orchestrates multiple AI agents in parallel and reported benchmark accuracy of 92% with o1, compared with 68% for out-of-the-box retrieval-augmented generation on its cited benchmark. ### Was the event information consistent across sources? (hebbia.com) A GarysGuide listing for the same May 20 event matched the Luma page on the key speaker names, topics and timing, including Osemwegie at Hebbia and Sosa at Noemica. Fonzi AI’s own event page, however, showed a variation. That page listed Sebastian Sosa as “Member of Technical Staff, @ Veris AI” rather than founder of Noemica, while keeping the rest of the lineup and agenda substantially similar. (openai.com) The Luma and GarysGuide listings both identified Sosa with Noemica. ### What happens next? (garysguide.com) The May 20 event page said registration was required and marked the gathering as part of Fonzi’s New York AI Engineers series. Fonzi described itself on the event listing as a recruiting platform focused on product and AI engineering hires in New York and San Francisco, with “75+ open roles” across startups in those cities. (luma.com) (fonzi.ai)