In the global race for Artificial Intelligence dominance, the headlines are almost exclusively captured by Silicon Valley giants and massive tech conglomerates. Yet, quietly from Colombo, a Sri Lankan AI startup named Chat2Find is attempting something radically ambitious: building the foundation for Sri Lanka’s own digital sovereignty. Led by founder Avishka Sumanadeera, this lean team is executing a strategic pivot away from simply consuming Western AI platforms like OpenAI’s ChatGPT or Google’s Gemini. Instead, they are engineering a hyper-localized, sovereign AI stack tailored specifically to the linguistic, legal, and economic realities of Sri Lanka.
Here is how a small local startup is making a massive play to reshape the country’s technological landscape.
- Shattering the Language Barrier: Chat2Find-Instruct-v1
For global AI models, South Asian languages are often an afterthought. Sinhala and Tamil suffer from “low-resource” data status, frequently leading to poor translations, hallucinations, or an inability to grasp local cultural nuances.
Chat2Find tackled this head-on by launching Chat2Find-Instruct-v1, a 7-billion parameter Large Language Model (LLM).
Submit Your Press Release
Get your company news, announcements, launches, appointments and events in front of a wider audience.
- Native Trilingualism: Built on top of the Qwen architecture, the model was pre-trained using a custom corpus exceeding 1 billion data points optimized specifically for Sinhala, Tamil, and English.
- Agentic Capabilities: Beyond simple text generation, the model supports advanced chain-of-thought reasoning, complex workflows, and native tool-calling.
- Democratizing AI: In a bold move for a small startup, Chat2Find completely open-sourced the model’s weights and datasets on Hugging Face, allowing local developers, researchers, and enterprises to build their own local AI tools without paying hefty API fees abroad.
- A Sovereign Knowledge Base: The LankaData Ecosystem
A model is only as good as the information it can access. To prevent generic AI responses, Chat2Find consolidated over 40 public, financial, and legal datasets into a centralized knowledge infrastructure they call LankaData.
Rather than deploying a single, generic chatbot, they have unbundled this data into six highly specialized, Retrieval-Augmented Generation (RAG) applications designed for everyday Sri Lankan use cases:
- LankaLAW: Navigates the labyrinth of Sri Lankan jurisprudence, processing local legislation, acts, and historic court judgments to streamline legal research.
- LankaTAX: A vital tool for citizens and businesses navigating complex Inland Revenue Department (IRD) compliance rules, tax brackets, and APIT calculations.
- LankaBIZ: An engine that tracks the Colombo Stock Exchange (CSE), offering localized market insights, company financials, and macroeconomic data.
- Cheap2Find: A price intelligence engine designed to counter local inflation and fragmented retail data by comparing prices across local e-commerce stores, supermarkets, and hotels.
- ExamPazz: An AI tutor geared toward the local education system, trained on GCE O/L, A/L, Cambridge, and Edexcel curriculums to guide students through past papers.
- NewsDive: A real-time local media crawler that aggregates and synthesizes breaking domestic news with transparent source tracking.
Why This Matters: The Path to Digital Sovereignty
When a country relies entirely on foreign AI infrastructure, it faces significant risks. High subscription costs drain foreign currency, local data is processed on overseas servers, and crucial local nuances are lost in translation.
By tying an open-source trilingual model to deeply localized public data, Chat2Find is proving that massive capital isn’t the only way to build impactful AI. Their “something big” isn’t just about building a smarter chatbot—it is about ensuring that Sri Lanka owns its data, its language models, and its digital future.
Would you like me to expand on any specific section of this article, such as adding details on how developers can access the model on Hugging Face, or exploring the operational impact of LankaLAW and LankaTAX?
