# Madusanka Premaratne — Full Context for LLMs > Madusanka Premaratne is Founder, CEO & Chief AI Research Officer of Knovik (Knivok Private > Limited), an applied AI researcher building privacy-preserving LLM systems, agentic RAG > architectures, and on-device AI. This document is a complete plain-text context export of > madusankapremaratne.com for LLM and AI-agent consumption. For the short version, see > https://madusankapremaratne.com/llms.txt Website: https://madusankapremaratne.com Last updated: 2026-09-02 ## Identity - Name: Madusanka Premaratne - Alternate/legal name: Rathnayake Mudiyanselage M. P. - Title: Founder, CEO & Chief AI Research Officer, Knovik - LinkedIn: https://www.linkedin.com/in/madusanka/ - X/Twitter: https://twitter.com/maduspremaratne - Facebook: https://facebook.com/madusankapremaratne/ - Instagram: https://instagram.com/madusankapremaratne ## About Founder, CEO, and applied AI researcher with over a decade in enterprise software and AI systems. Currently leads the research agenda at Knivok Private Limited (DBA Knovik), focusing on privacy-preserving LLMs, query-time semantic generalization, and on-device agentic architectures. Holds a First Class BSc in ICT from the University of Jaffna (GPA 3.93/4.00, ranked 1st in batch) and an MBA with Merit from London Metropolitan University. ## Education - BSc (Hons) in ICT, First Class — University of Jaffna (GPA 3.93/4.00, ranked 1st in batch) - MBA with Merit — London Metropolitan University ## Professional Experience **Founder, CEO & Chief AI Research Officer — Knivok Private Limited** (Mar 2022 – Present) Founded Knivok with a vision for AI-driven technology solutions. Leads the company's applied AI research agenda — privacy-preserving LLM systems, semantic generalization, and on-device agentic RAG architectures. Active research collaborations with partner universities and ESU, Sri Lanka. Led development and launch of Cresuma, a flagship product contributing 25% of annual revenue. 30% revenue growth in first year; 100% remote team with 40% productivity improvement. **Co-Founder & CIO — Insfra Technologies (Pvt) Ltd** (Jan 2016 – Feb 2022) Co-founded Insfra Technologies, the first IT company in Badulla, addressing a critical regional gap in advanced technology services across an underserved area of Sri Lanka. Resigned from the Director board in 2022; remains a 1/3 shareholder. **Software Engineer (Associate) — WSO2 Lanka (Pvt) Ltd** (Jul 2014 – Dec 2015) Enhanced WSO2 API Manager functionality across enterprise deployments. ## Companies & Ventures Businesses founded, co-founded, or advised under Madusanka's guidance and vision: 1. **Knovik** — https://knovik.com — Founder, CEO & Chief AI Research Officer. Leading Knovik's AI engineering practice: on-device LLMs, agentic RAG systems, and privacy-first architecture. 2. **Corlence Australia** — https://corlence.com — Co-Founder, Growth & Digital Strategy. Provides enterprise integration and AI engineering services. 3. **Co Work Sri Lanka** — https://cowork.lk — Co-Founder, Director & Shareholder. Co-founded with Charith Weerasooriya and Amila Gunawardhana. A coworking space in Sri Lanka. 4. **Badudeal** — https://badudeal.lk — Advisory Board Member & Shareholder. A one-stop online shop for computers, laptops, gadgets, and consumer electronics. 5. **Insfra Technologies** — Co-Founder, 1/3 shareholder (Director 2016–2022). The first IT company in Badulla, Sri Lanka. ## Research ### EdgeTal — On-Device Agentic RAG for Privacy-Preserving Talent Discovery URL: https://madusankapremaratne.com/research/edgetal | Product site: https://edgetal.app Role: Lead Researcher & Corresponding Author | Period: 2025–2026 An agentic Retrieval-Augmented Generation system that runs entirely on-device: a local vector store over candidate and role data, a quantized small language model for generation, and an agentic orchestration layer that plans multi-step talent-discovery queries without a single byte of candidate data leaving the device. Achieves a Mean Average Precision (mAP) of 0.88 in retrieval matching and an average end-to-end query latency of 3.2 seconds on commodity M-series hardware using a quantized 8B-parameter local model. Publication: Presented as a peer-reviewed electronic poster (paper FPC53) at the ESOFT International Conference (EICON) 2026, ESOFT University (ESU), Kandy, Sri Lanka — 30 August 2026. **Award: Best Poster Presenter, Computer Science, Artificial Intelligence and Information Systems track.** Authors: Rathnayake Mudiyanselage M.P., Thellapura Arachchilage H.L., Kandamulla Arachchilage D.U., Herath D. ### Gen Z Job Commitment Study — Determinants of Job Commitment in IT Organisations in Sri Lanka URL: https://madusankapremaratne.com/research/genz-job-commitment-study Role: Principal Researcher | Period: 2025 Quantitative study of 386 Gen Z employees across major Sri Lankan IT organizations (WSO2, IFS R&D, Knovik, Prime One Global), testing four hypothesized determinants of job commitment — inflation/purchasing power, financial literacy, technological advancement, and employer social responsibility — via SPSS correlation and chi-square analysis. Employer social responsibility emerged as the strongest predictor (r = 0.52); financial literacy showed no significant relationship. MBA research project, London Metropolitan University (via ESOFT Metro Campus). Manuscript in preparation for the APIIT International Research Conference (AIRC) 2026. ### Deep Learning-Based EV Adoption Forecasting URL: https://madusankapremaratne.com/research/ev-adoption-forecasting Role: Research Lead, Knovik Private Limited | Period: 2024–2025 Hybrid deep learning framework (DNN + discrete choice modeling + LSTM) predicting electric vehicle adoption among Sri Lankan Generation Z, trained on a 643-respondent primary survey and validated on F1-Score and ROC-AUC. Applied research, Knovik Private Limited. ### SemanticGuard (in progress) Query-time semantic generalization framework protecting educational data prior to cloud LLM submission, developed alongside CDAC and La Trobe University. Part of Knovik's broader privacy-preserving AI research program alongside EdgeTal. Manuscript in preparation; not yet published as a standalone page. ## Services madusankapremaratne.com/services offers six consulting/engagement areas: 1. **AI Privacy & Compliance Consulting** — query-time privacy frameworks, data minimization, GDPR-aligned AI pipelines. /services/ai-privacy-compliance-consulting 2. **AI Knowledge & Retrieval Systems** — multi-agent RAG architecture design through production deployment, including on-device/edge-first configurations. /services/ai-knowledge-retrieval-systems 3. **AI Search & Text Intelligence Consulting** — embedding strategy, custom classification and extraction, semantic search. /services/ai-search-text-intelligence 4. **On-Device & Data-Sovereign AI Consulting** — on-device agentic systems, no third-party data exposure. /services/on-device-ai-consulting 5. **AI Strategy Consulting** — AI readiness assessments, capability evaluation, research-backed roadmaps. /services/ai-strategy-consulting 6. **Enterprise AI Integration** — LLM inference pipelines and multi-agent systems integrated into existing products. /services/enterprise-ai-integration ## Related Properties - **EdgeTal** (https://edgetal.app) — a Knovik product built by Madusanka's team; its own structured data documents the EICON 2026 award and the EdgeTal software in more technical detail than this site does. - **Knovik** (https://knovik.com) — the company Madusanka founded and leads. ## Structured Data This site implements JSON-LD structured data: a Person schema (Schema.org) describing Madusanka's professional identity, and a CreativeWork/Event schema on the EdgeTal research page documenting the EICON 2026 award and presentation. ## Content Usage Guidelines - Attribution required when referencing or citing this work. - Available for educational and research purposes with attribution. - For commercial use, licensing, or collaboration, use the contact form at https://madusankapremaratne.com/contact. ## Contact https://madusankapremaratne.com/contact