-
-
Meritt-AI
Inspiration
In 2020, during the #EndSARS protests, I witnessed firsthand how many Nigerians — especially youth — were arrested without reason, often jailed for years without trial. I saw businesses looted, families devastated, and ordinary citizens confused about their rights. Police brutality was rampant, yet many didn't know the law was on their side — or how to access it. The legal system felt like a fortress built to keep justice away from the people.
As a software engineer and founder, I decided to build something for everyone — individuals, lawyers, judges, even the police — to help them understand, act on, and deliver justice efficiently. That vision became Meritt: an AI-powered legal companion for Nigeria.
What it does
Meritt is a legal AI assistant designed to give everyday Nigerians — and legal practitioners — instant access to trusted legal information, advice, and tools. It helps individuals resolve daily legal issues (e.g., police harassment, tenancy disputes, workplace injustice), enables lawyers to research and draft faster, and supports judges or legal officials with better access to precedents and document digitization. All of this is delivered via a simple interface: chat, voice, or embedded tools.
How we built it
- A Retrieval-Augmented Generation (RAG) system for Nigerian laws (federal, state, Sharia, customary).
- Supabase as the vector store and content backend.
- A frontend built with modern JavaScript (Svelte/React).
- Integrated with LLaMA-3 via AwanLLM for reasoning and custom prompts.
- Scrapping the internet, NGO collaborations and many others
- PDF ingestion, smart legal document parsing, and AI-powered answer generation.
Challenges we ran into
- Gathering and structuring Nigerian legal data across different law systems (federal, state, customary, Sharia) from fragmented sources.
- Hosting, Graph RAG setup and other Costs.
- Optimization of Agent Q for Autonomous tasks and Internet related tasks
- Aligning legal AI generation with jurisdiction-specific rules and interpretations.
- Managing latency and cost while maintaining accurate, real-time results for users.
- Ensuring ease of use for a very diverse audience — from rural users in need of legal help to high-performing legal teams.
- Working with NGOs and Remands ## Accomplishments that we're proud of
- Deployment/hosting of the Frontend, Backend and Llama Model
- Init Setup for proper RAG and later Fine turning depending on the result.
- Autonomous setut for find and scrapping latest law judgement and report across the internet
- Created a single legal research interface usable by both laymen and legal professionals.
- init/pipeline Agent Q setup
- Three Lawyer firms tested the init RAG and gave positive feedback
What we learned
- Many Nigerians have legal needs that go unmet due to cost, access, or fear of institutions.
- Most disputes (landlord-tenant, police, employment) occur due to ignorance of the law — not intent.
- Lawyers, especially solo practitioners, struggle with access to updated and structured legal documents.
- Embedding AI into legal workflows can drastically reduce research and drafting time.
- This could go beyond the legal industry
What's next for Meritt
- Partner with legal aid groups, courts, NGOs, and law firms for deeper integrations.
- Add support for personalized legal profiles (case tracking, reminders, history).
- Expand state-level law ingestion and support indigenous languages.
- Get Full RAG, possibly Fine turning, and Agent Q to full development
- Begin fine-tuning on Nigerian case law data and incorporate precedent prediction and judgment analysis.
Log in or sign up for Devpost to join the conversation.