Below is a Devpost / hackathon–ready write-up, written directly from your architecture and pipeline, techy but clear, and easy for judges to scan. You can paste this as-is.
💡 Inspiration
Immigration information is scattered across government websites, forums, and outdated blogs. People spend hours researching, still unsure if they are missing a step or risking their legal status. We wanted to build a system that acts like a real-time pulse of immigration rules—constantly monitoring official sources and turning complexity into clarity.
⚙️ What it does
ClarityPulse is a real-time immigration roadmap generator.
Given a user’s current visa, target visa, and basic personal context, it:
- Generates a step-by-step immigration roadmap
- Calculates timelines, deadlines, and total duration
- Breaks down government fees, legal costs, and documents
- Highlights decision points with recommendations
- Identifies risks and when to consult an immigration attorney
- Provides citations from official government sources
All outputs are structured, traceable, and ready to use.
🏗️ How we built it
We built ClarityPulse as a parallel AI pipeline using three AI services:
- OpenAI (Search Strategy Generator) Understands the user’s immigration scenario and generates:
- Targeted search strategies
- Relevant government websites
- Key legal and procedural terms
- Parallel Data Gathering
- Perplexity retrieves structured, up-to-date immigration data with citations
- Firecrawl scrapes official government websites (USCIS, DOL, State Dept) in parallel
- OpenAI (Final Synthesis Engine) Combines all sources into a single, comprehensive immigration roadmap, including timelines, costs, decisions, and legal guidance.
The system outputs four structured JSON files, with a final production-ready roadmap as the main artifact.
🧗 Challenges we ran into
- Fragmented and inconsistent immigration data
- Balancing authoritative sources with recent web information
- Ensuring structured, deterministic outputs from LLMs
- Parallelizing external API calls while keeping latency low
- Avoiding legal hallucinations and ensuring proper citations
🏆 Accomplishments that we're proud of
- Built a fully parallel, end-to-end AI pipeline
- Combined official government data + recent web intelligence
- Generated actionable outputs (dates, costs, decisions—not summaries)
- Included explicit attorney consultation triggers
- Achieved ~40 seconds end-to-end execution time
- Designed outputs that are UI- and production-ready
📚 What we learned
- Parallel AI workflows dramatically improve both speed and reliability
- Structured outputs are critical for trust in legal and compliance domains
- Combining multiple AI services produces far better results than a single model
- Clear decision points matter more to users than raw information
- Citations are essential for credibility in high-stakes domains
🚀 What's next for ClarityPulse
- User dashboards to track immigration progress over time
- Live alerts for rule changes, deadlines, and processing times
- Country-agnostic expansion beyond the U.S.
- Attorney and advisor integrations
- Personalized scenario simulations (best-case vs worst-case paths)
ClarityPulse aims to become the real-time heartbeat of immigration journeys, replacing confusion with clarity.
Built With
- apifirecrawl
- apis
- cli-based
- data
- dotenv
- engineering
- json
- openai
- parallel
- perplexity
- pipeline
- processing
- prompt
- python
- rest
- validation
Log in or sign up for Devpost to join the conversation.