MedFind
Healthcare Platform
AI/ML2026
VizaCheck
Immigration Tech
Immigration agencies were spending hours manually reviewing applicants' social media presence—a process that was slow, inconsistent, and missed critical patterns. I built an AI-powered screening system that analyzes digital footprints across platforms, detects concerning patterns in multiple languages, and generates compliance-ready reports. The system now processes 10K+ applications monthly.
Company
Thelix
Role
Full Stack Engineer
Duration
8 months
Year
2026
Tech Stack
Next.jsNode.jsPythonAI/MLPostgreSQLRedis
The Problem
Manual social media reviews took 2-3 hours per applicant, created inconsistent results across reviewers, and couldn't scale with application volume.
The Solution
Built an AI pipeline that ingests social media data, runs multi-language NLP analysis, and generates risk assessments with audit trails—reducing review time to minutes while improving consistency.
Technical Challenges
1
Designing data pipelines that could process thousands of social posts without hitting rate limits
2
Building NLP models accurate across English, Spanish, and Mandarin content
3
Implementing GDPR-compliant data handling with full audit trails
4
Creating explainable AI outputs that immigration officers could trust and verify
Results & Impact
90%
Processing Time
reduction in review time
95%+
Accuracy
detection accuracy
10K+
Scale
applications processed monthly
Lessons Learned
Key Takeaways
- AI systems need explainability—users won't trust black boxes for high-stakes decisions
- Rate limiting and backoff strategies are essential when aggregating external data
- Audit trails aren't optional in regulated industries