AI Deploy Network
AI AgentCapital Markets 11/16/2026

Nairacle: AI Research Agent for the Nigerian Stock Market

Nairacle is an AI research agent for the Nigerian stock market. It combines financial news retrieval, live stock prices, and historical market data to answer company research questions with supporting sources.

Hours Automated
0
Cost Savings
Currency not specified
Revenue Impact
Currency not specified

Business Challenge

Researching companies listed on the Nigerian Exchange can involve reading dozens of articles, checking stock prices, and comparing information published at different times. General-purpose search tools often return outdated news alongside recent developments. This makes it difficult to understand what is happening with a company without spending time checking several sources. I wanted to build a research assistant that could collect relevant financial information automatically and help users investigate Nigerian companies through natural-language questions. The main challenge was handling large amounts of financial news while keeping retrieval focused on relevant, recent information and providing sources users could check.

Solution Delivered

I built and deployed Nairacle, an AI research agent for the Nigerian stock market. The system runs a daily ingestion pipeline that collects financial news, classifies articles for relevance, summarizes useful content, generates embeddings, and stores the results in Pinecone. I used Python, LangChain, LangGraph, Google Gemini, and Tavily to build the research workflow. When someone asks a question, the agent searches stored information by meaning, considers article recency, and reranks the results before producing an answer with supporting sources. The retrieval system uses a default six-month recency window that can adjust to the user's question. I also integrated TradingView Screener for market prices and a PostgreSQL database hosted on Neon for historical financial data. LangSmith records agent execution and tool calls, which helps me inspect how answers are generated and investigate failures. The application has a web interface connected to its Python backend through REST APIs and is deployed on Hugging Face Spaces.

Outcomes Achieved

Nairacle makes Nigerian stock market research accessible through a conversational interface. Instead of manually searching for articles and reading them individually, users can ask questions about listed companies and receive responses supported by retrieved financial news and market data. The automated ingestion pipeline keeps adding new financial information without requiring me to collect and process each article manually. The system also improves how the research data is organized. Relevant articles are classified, summarized, embedded, and made searchable through a single application. I integrated LangSmith tracing to inspect retrieval and agent behaviour, making it easier to identify errors and improve the system. The current deployment demonstrates automated financial data processing, conversational research, source retrieval, and integration with structured market data.

Measurable Business Outcome

The financial news ingestion pipeline processed approximately 14,000 articles and identified about 4,900 relevant articles for the research system. The pipeline runs automatically each day, collecting and processing new financial information without manual article-by-article handling. Users can ask natural-language questions and receive sourced research responses in seconds. The retrieval system also applies a default six-month recency window, which adjusts when a question requires a different period. These figures describe the application's processing capacity and research workflow. I have not yet measured customer-level time savings, revenue impact, or financial returns.

Business Outcome Categories

Knowledge ManagementTime SavingsProductivity ImprovementQuality Improvement