endorsedCompleted

MEWA Chatbot: AI-Powered Electoral-Procedure Assistant

Services Covered:

Chatbot Development • AI and Machine Learning

Industry

Education

Duration

7 months

Budget

10K - 49K

Start Date

1 February 2025

Team Size

1-5 Employees

Engagement Model

Onshore

Project Capability Score (PCS)

77/100

Good

Project Capability Score (PCS) estimates how capable nomtek is of successfully delivering a project like this, based on its past experience, track record, and ability to handle similar work. Learn more about PCS

Client

The Center for Citizenship Education

Warsaw, Poland

Project Summary

Nomtek designed and built MEWA, an AI-powered chatbot for a Polish nonprofit, implementing a Python-based backend integrated with LlamaIndex and the OpenAI LLM stack and a vector database. They ran a Figma discovery workshop, delivered a query-routing engine, set up automated evaluations for factual correctness, and provided cost-optimization strategies while running efficient two-week sprints and close Slack communication.

Key Challenges

Build an AI-powered chatbot that gives voters verified, unbiased answers drawn from the Center for Citizenship Education knowledge base.
Integrate the chatbot seamlessly into the Latarnik Wyborczy (Electoral Lighthouse) platform.
Ensure factual correctness and transparency in responses.

Project Deliverables

Backend for MEWA chatbot (Python, LLM integration)
Vector-database integration for curated electoral documents
Discovery workshop (Figma)
Query-routing engine (classification, identity and hate-speech detection)
Automated evaluations for faithfulness and factual correctness
Cost-optimization strategies (caching, token trimming, future self-hosted models)

Project Solution

Nomtek designed and built the MEWA chatbot backend using Python, LlamaIndex, and the OpenAI LLM stack, connecting it to a vector database of curated electoral documents. They ran a Figma discovery workshop, implemented a query-routing engine (classification, identity and hate-speech detection), set up automated evaluations for faithfulness and factual correctness, and prepared cost-optimization strategies including caching and token trimming.

Project Outcome

Working MVP answers electoral-procedure questions with >92% factual-correctness in internal evaluations.
Positive pilot feedback from >100 student testers; 87% said MEWA helped clarify voting rules.
Delivered every milestone on time and ran crisp two-week sprints with daily Slack check-ins.

Platforms

Tech Stack

Python

Client Endorsement

Overall Review Rating

Timeliness

Cost Rating

Willing to Refer

Quality of Deliverables

“nomtek has delivered a functional MVP that answers questions with over 92% factual correctness, and the client has received positive feedback from over 100 student testers. The team is proactive and solution-oriented. nomtek runs efficient sprints and collaborates via Jira and daily Slack check-ins.”
Anonymous
Project Manager
Date: 07/28/2025