DEV Community: Mustapha Alaba

AI-driven Conversational analytics from MongoDB Health Insurance records

Mustapha Alaba — Tue, 03 Mar 2026 23:39:07 +0000

This is a submission for the Built with Google Gemini: Writing Challenge

What I Built with Google Gemini

Our client is a digital health insurance company serving 26,000+ members across East Africa. Their Relationship Managers and HR partners had no way to query their own data without routing requests to engineers — meaning answers to routine questions like "How many wellness visits did Company X complete this quarter?" took hours or days. I lead my practicum team to build an AI-driven conversational analytics platform that lets non-technical staff ask questions in plain English and get instant responses — as tables, charts, and plain-English summaries — tested on development MongoDB data.

Architecture: Classify → Generate → Execute
The core insight was that sending a raw natural language query straight to an LLM and hoping for accurate MongoDB Query Language (MQL) output is unreliable in a healthcare context. So we split the work into two stages:

Intent Classification — Gemini classifies the query into one of 14 predefined intent types (e.g., aggregate_one, filter_lookup, geographic_breakdown)
MQL Generation — Gemini generates the aggregation pipeline with the right prompt template and schema context injected for that specific intent

Demo

What I Learned

The main takeaway from this project is the superpower that LLMs like Gemini now give us as builders. What would traditionally require weeks of work — designing a custom analytics dashboard, assembling a dedicated team to maintain it, and hardcoding queries for every use case — was reduced to just the time needed to integrate Gemini with the relevant database collections. The system is also dynamic: it adapts automatically to new data structures, and the active collections can be expanded or reduced with just a few clicks.

Another important lesson is to build guardrails and logic that guide the LLM on what it can do and what it has access to. In this project, that was achieved through a configuration page where administrators can select which collections and fields Gemini is permitted to query. The existing authentication and authorisation infrastructure was also leveraged to enforce which users can access specific query results. The value of audit trails became very clear throughout the process — logging LLM behaviour and query outputs made it significantly easier to track errors, identify failure patterns, and continuously improve the overall system.

Teamwork — both within our practicum team and with the client's technical team — was critical to the progress we made. We set up weekly meetings for feedback and testing with the system's intended end users, which made it straightforward to quickly refine our approach and align closely with the client's requirements.

Finally, AI tools like Gemini are genuinely powerful — but only when used with intention, structure, and the right guardrails in place.

Google Gemini Feedback

With the right level of context and guidance for Gemini, the accuracy of the generated queries increased significantly. We initially started with no context, and the generated queries were more of a gamble than a reliable output. After setting up an intent analyser and passing the table schemas to improve context, the results improved significantly.

Where we hit friction: Date handling was the biggest pain point. Gemini generated date filters that looked correct but failed silently — returning zero results because it used JavaScript Date string literals instead of MongoDB ISODate objects. We had to encode this as an explicit rule in every prompt and add a post-generation validation layer.

Multi-collection queries hit an accuracy ceiling too. Single-collection queries exceeded 85%, while multi-collection $lookup queries dropped noticeably as injecting context for multiple schemas increased prompt ambiguity. Semantic retrieval via a vector database is the planned fix.

Diascora - Connects Diaspora Communities in Rwanda

Mustapha Alaba — Mon, 02 Mar 2026 07:36:58 +0000

This is a submission for the DEV Weekend Challenge: Community

The Community

Rwanda has become a growing hub for higher education and remote work in Africa. Through partnerships with institutions such as Carnegie Mellon University and African Leadership University, the country attracts students from across the continent. In addition, Rwanda’s stable environment draws freelancers and professionals who relocate temporarily or permanently.

One active sub-community within this ecosystem is Nigerians in Rwanda—a network of students, working professionals, and entrepreneurs. Members of this community frequently need peer-to-peer support in two key areas:

Currency Exchange: Converting between the Nigerian Naira and the Rwandan Franc. Existing external services often result in significant losses due to high transaction fees and unfavourable exchange rates.

Item Transfers: Coordinating with trusted travellers to carry small but important items—such as documents, certificates, or hard-to-find goods—between Nigeria and Rwanda.

Currently, these requests are managed informally through multiple WhatsApp groups, making coordination fragmented, inefficient, and less secure.

To address these challenges and better serve the community I belong to, I decided to build a dedicated platform that streamlines peer-to-peer exchanges and item transfers in a more structured, transparent, and cost-effective way.

What I Built

Diascora is a web platform for diaspora communities to coordinate peer-to-peer currency exchanges and package deliveries — replacing scattered WhatsApp group messages with a structure and transparency.

The app has two core features:

Exchange Board
Members post currency exchange requests specifying the pair (e.g. NGN → RWF), amount, offered rate, and payment method. Other members browse the board and express interest. The poster reviews all interested peers, accepts one, and the contact details are revealed to both parties. Live exchange rates are fetched and cached to show how
A posted rate compares to the official market rate, helping both sides make informed decisions.

Delivery Board
Members who need an item carried between Nigeria and Rwanda post a delivery request with destination, package weight, a payment offer, and a description of what needs to be carried. Travellers who see a request they can
fulfil submit a carry offer. The poster accepts the best offer, which triggers contact reveal for both parties and notifies rejected travellers automatically.

Notifications are delivered both in-app and as browser push notifications (PWA).

Demo

diascora-o9nilpos.on-forge.com

Code

alabamustapha / diascora

How I Built It

NB: ClaudeAI was used

Backend: Laravel 12 on PHP 8.4, following the streamlined Laravel 11+ application structure. All reactive UI is handled server-side with Livewire 4 — no separate API layer needed. Business logic lives in Livewire component actions, with Form Request validation and Eloquent relationships keeping controllers thin. Database transactions wrap the accept-offer lifecycle to ensure atomicity.

Frontend:
Flux UI v2 (the official Livewire component library) provides all form controls, modals, badges, and navigation components. Tailwind CSS v4 handles layout and a custom navy brand palette. The app is a PWA — a service worker handles push notification delivery and notificationclick routing.

Auth:
Laravel Fortify for email/password auth and Laravel Socialite for Google OAuth. First-registered user is automatically promoted to sysadmin. Role management uses Spatie Laravel Permission v7.

Notifications:
database (for the in-app bell) and laravel-notification-channels/webpush for browser push.

Queues & Observability:
Laravel Horizon manages the Redis-backed queue with a dashboard restricted to sysadmins.

Laravel Telescope is available in development for inspecting requests, queries, and jobs.

Testing: Pest v4 with 100+ feature tests covering board access, filter logic, the full offer/interest lifecycle, notification dispatch (Notification::fake()), file uploads (Storage::fake + UploadedFile::fake()), and double-submit guards.

Dev environment: Docker via Laravel Sail, with SQLite in-memory for the test suite to keep test runs fast without a running database container.

Hosting:
Server management using Laravel Forge with a VPS server from Digital Ocean

AI Taxi Dispatch Operator

Mustapha Alaba — Sun, 23 Jun 2024 22:57:27 +0000

This is a submission for Twilio Challenge v24.06.12

What I Built

Optimizing Business Efficiency with AI Operators
Many businesses spend hours daily taking orders over the phone from prospective clients, leading to less productive employee time and increased operational costs.

AI Operators can revolutionize this process by intuitively collecting orders from prospective clients and notifying the business through email or external APIs when an order is ready for fulfillment. This automation will enhance employee productivity and reduce business costs by streamlining the manual order collection process.

A compelling proof of concept for such AI Operators is the AI Taxi Dispatch Operator. This solution enables taxi companies to automate the booking process and dispatch drivers and vehicles using the company's dispatch system via APIs or manual assignments.

In this implementation, Twilio WhatsApp messaging is utilized to create an AI operator built with OpenAI technology. The AI operator gathers all necessary information from the user, sends out dispatch requests through the company's dispatch API, and notifies both the company and the user when a driver has been dispatched.

By integrating AI Operators, businesses can significantly improve efficiency, reduce costs, and provide better service to their clients.

Demo

Github Repo

alabamustapha / twilio-taxi-operator

This is a submission for Twilio Challenge v24.06.12

What I Built

Optimizing Business Efficiency with AI Operators Many businesses spend hours daily taking orders over the phone from prospective clients, leading to less productive employee time and increased operational costs.

In this implementation, Twilio WhatsApp messaging is utilized to create an AI operator built with OpenAI technology. The AI operator gathers all necessary information…

View on GitHub

Youtube Demo and Screenshots

Demo 1

Demo 2: App and Explanation

Twilio and AI

Twilio WhatsApp Messaging was used as the primary communication medium for the users. Messages sent to the WhatsApp number are processed through webhook by a web app built on Laravel. These incoming messages are sent to an AI assistant powered by the OpenAI GPT4 model. The AI assistance is tasked to handle the conversation until all required details to request a booking are provided after which the booking details is sent to the database for dispatching followed by email notification to admin and dispatch notification to the user with drivers details

Additional Prize Categories

Twilio Times Two

I used the Twilio WhatsApp Messaging feature for the interaction between the user and the operator. Twilio SMS Feature is used for failover notifications.

Impactful Innovators

This product will help companies reduce operation costs, save time, and increase lead generation.