πŸ† HackPrinceton'26 Β· Team Project

ACTA AI

Full-stack multi-agent contract intelligence platform for Clinical Trial Agreements

"50 million patients are waiting to enroll in clinical trials β€” not because the science isn't ready, but because lawyers are still negotiating paperwork."

100

days avg. to finalize CTAs

$1.4M

lost pipeline value / day

400K+

active clinical trials globally

50M

patients waiting for trials

The Problem

Clinical Trial Agreements take an average of 100 days to finalize β€” costing $800K–$1.4M in lost pipeline value per day. Lawyers negotiate the same clauses across the same contract types, manually, every time. For Regeneron alone with 56 active programs, cutting negotiation time by 50 days unlocks $2.24 billion in pipeline value β€” and gets patients into trials two months sooner.

Architecture

4-Stage LangGraph Pipeline

πŸ“„STEP 01

Document Parser

Detects headings with 3 regex strategies + DOCX style metadata

πŸ”STEP 02

Clause Extraction

CTA-specific taxonomy: Publication Rights, Subject Injury, Protocol Deviations

⚠️STEP 03

Risk Identification

Hybrid rule-based + LLM analysis with RED/YELLOW/GREEN scoring

βœ…STEP 04

Compliance Check

Flags missing required clauses as compliance findings

✍️STEP 05

Suggestion Generator

Produces negotiation-ready redlines with git-style diffs

Smart Model Router

Right model. Right job. Every time.

K2 Legal

Indemnification & IP clauses

Deep legal reasoning

Gemini 2.5 Flash

Fast classification

High-speed clause sorting

Claude Sonnet

Rewrite tasks

Negotiation-ready language

Built with

Tech Stack

🐍 Python FastAPIπŸ•ΈοΈ LangGraphβ–² Next.js 15πŸ”· TypeScript✨ Gemini 2.5 FlashπŸ€– Claude Sonnetβš–οΈ K2 Legal Model

Built in 36 hours.

We didn't sleep much. We have no regrets.

"The patients can't wait."