platoseedplatoseed
Verification & alignment infra for AI decisions. MIT / BU team, design partnerships with Accenture, Dropbox, and Vercel.
Modaic is building the verification and alignment layer for AI decisions. For judgment-heavy workflows like classification, moderation, and automated evaluation, teams today are stuck choosing between expensive human review and weak automated systems that constantly break on edge cases. Modaic solves this by scoring how confident a model is in each decision, routing only uncertain cases to humans, and using that feedback to continuously improve it's underlying instructions. The result is decision automation that gets more reliable over time while dramatically reducing the amount of manual oversight required to run it.
Modaic provides verification and alignment infrastructure to help scale large language models for tasks like classification, ranking, judging, and extraction. It emphasizes aligning AI decisions with desired outcomes and offers tooling to improve consistency across model outputs.
Modaic offers an infrastructure layer to verify and align AI decisions. It supports scaling LLMs for various tasks (classification, ranking, judging, extraction, and more) by providing tooling to ensure outputs align with specified criteria and reduce misalignment across model predictions.
Who it’s for: Organizations deploying large language models who need improved decision alignment and verification across classification, ranking, judging, and extraction tasks.
partnerships with design collaborators (Accenture, Dropbox, Vercel) and academic team (MIT / BU)
Hey, I'm Farouk! My interests lie in AI-human interaction & search. Prev. Wayfair, IBM, Harvard
MIT '25 AI+D, prev @ MIT Media Lab, Bloomberg, Axon - Trained foundation transformer models at the MIT Media Lab
The monitoring layer your AI agents need.
Ex. OpenAI, Palantir, Sandia Labs building the AI operating system for consulting firms
