Goal
- Ship a WhatsApp Service for KIND clients: Q&A + issue escalation, 24/7, multilingual.
- Constrained cost-per-user and knowledge-base using LLM techniques.
- Capstone window: Jan–May 2024, 5-person team.
Back to projectsN/ALive demo under NDA.
ContigoTechnical Project Manager + SWE
Role
I was the technical project manager and a software engineer on a five-person UW iSchool capstone team (Jan–May 2024), sponsored by Kids in Need of Defense (KIND). I guided our development process regarding AI use, designed the technology stack based off client needs, and wrote the Flask application. The MVP was shipped and designed to fufill the needs of 10 clients and their case-managers as a proof-of-concept.
Problem
Navigating asylum and legal services is high-stakes for Unaccompanied Alien Children. KIND needed a low-cost, multilingual channel that answers questions based off their written existing knowledge base, capable of escalating issues to case managers when appropriate, and can be institutionally adopted without signifigant staffing requirements.
Skills and technologies used
- Technical project management
- Retrieval-Augmented Generation
- LLM output evaluation
- WhatsApp Cloud API
- Python / Flask
- MongoDB
- Azure App Service
Overview
1 / 4
Demo
Capstone walkthrough of Contigo’s WhatsApp chatbot proof of concept for KIND.