
RegPilot AI is a dual-agent compliance copilot built for founders who can't afford $10,000/hour legal review or 3-6 months of manual regulatory research before shipping. Paste a startup idea, select a target region, and RegPilot AI retrieves the most relevant regulatory clauses using semantic search over Fireworks AI-generated embeddings, then runs two chained agents: a Compliance Agent that identifies specific gaps against GDPR, DPDP, CCPA, and the EU AI Act with severity ratings, followed by an Action Agent that converts those gaps into a prioritized remediation roadmap with immediate, 30-day, and 90-day action items. Every inference call in the pipeline — both the embedding generation and the LLM reasoning — runs on AMD Instinct MI300 GPUs through Fireworks AI, with zero local or third-party GPU usage. Built with FastAPI, Next.js, and a lightweight in-memory vector search layer. Fully deployed and live, not a local demo. Target market: startup founders, indie SaaS builders, and compliance-adjacent consultants navigating a $1B+ legal-tech gap with zero AI-native multi-region competitors.
13 Jul 2026

StudyBand is a multi-agent educational platform built for Track 1 (Internal Enterprise Workflows). It replaces the slow, manual process students go through to study a topic — researching, rewriting notes simply, creating practice questions, and checking answers — with four specialized AI agents that hand off work to each other automatically through Band.ai. The Researcher agent gathers structured study notes on any topic. It passes these to the Simplifier agent, which rewrites them in clear, education-level-appropriate language. The Quiz Master agent then generates multiple-choice questions from the simplified notes. Finally, the Evaluator agent grades the student's answers, gives encouraging feedback, and — if the score is below 80% — automatically triggers the Quiz Master to generate a shorter remedial quiz on the weak topics, creating a real feedback loop between agents rather than a one-way pipeline. All agent-to-agent communication happens inside a shared Band.ai room using @mentions, the same way a human team would hand off tasks in Slack — Band is the actual coordination layer, not a wrapper around a single LLM call. Built with Band.ai, Groq (Llama 3.3 70B for low-latency inference), AI/ML API (for switching between GPT-4o, Claude, and DeepSeek), LangGraph, Python, and Streamlit. Deployed live on Render with both the UI and all 4 agents running together. Beyond the hackathon, StudyBand has a clear path to revenue: a low-cost monthly subscription for individual students, white-label licensing to coaching institutes, or direct adoption by universities as an internal learning tool.
19 Jun 2026