The problem OrbitAI solves

AI-powered agentic finance protocol designed to solve the onboarding, usability, and execution friction in Web3,solves usability, fragmentation, and safety barriers in Web3 using AI agents that convert plain-English requests into on-chain execution

Challenges we ran into

AI Integration – API Rate Limits & Cost
One of our biggest hurdles was hitting rate limits and unexpected cost spikes while integrating our AI model calls. During testing and demo prep, repeated requests quickly ran into throttling, which broke the user flow. We got past this by adding request batching/caching where possible, introducing retry logic with exponential backoff, and trimming unnecessary calls so we only hit the API when it truly added value.

Training – Data Quality Issues
Our training data was messy and insufficient for the model to generalize well early on. We had to go back and clean, deduplicate, and augment the dataset, then iterate on preprocessing to get more consistent inputs. This significantly improved output quality once we retrained.

UI/UX – Scroll Animation Conflicts with Next.js
Our scroll animation library didn't play well with Next.js's rendering/hydration behavior, causing glitches and inconsistent triggers. We fixed this by carefully scoping animation initialization to client-side only (avoiding SSR mismatches), and adjusting how/when the library mounted relative to Next's lifecycle.