# Kavindu Ravishan Perera > Project Researcher at the University of Oulu and a PhD candidate in reinforcement learning. Builds conversational AI, retrieval augmented generation and AI agent systems for policymakers, researchers and everyday users, and studies how people and AI work together. This file is a plain-text map of https://www.kavinduravishan.com/ for language models and answer engines. Everything in it is taken from the pages themselves and was current in September 2026. The full text of every page is at https://www.kavinduravishan.com/llms-full.txt. ## Identity - Full name: Kavindu Ravishan Perera - Also known as: Kavindu Perera, Kavindu Ravishan (papers are published as Kavindu Perera) - Role: Project Researcher, University of Oulu - Doctoral research: Reinforcement learning, University of Oulu (just starting) - Country of nationality: Sri Lanka - Languages: English, Sinhala - Email: ravishankavindu [at] gmail [dot] com - Website: https://www.kavinduravishan.com/ - About page, with bios and photographs: https://www.kavinduravishan.com/about/ - Photograph: https://www.kavinduravishan.com/images/kavindu-ravishan-perera.jpg - ORCID: https://orcid.org/0000-0001-9621-3104 - Google Scholar: https://scholar.google.com/citations?user=X6T7ffIAAAAJ&hl=en - LinkedIn: https://www.linkedin.com/in/raviya/ - ResearchGate: https://www.researchgate.net/profile/Kavindu-Perera-15 - GitHub: https://github.com/kravishan - YouTube: https://www.youtube.com/@kavindu.ravishan.perera ## Bio Kavindu Ravishan Perera (published as Kavindu Perera) is a Project Researcher at the University of Oulu, where he is starting a PhD in reinforcement learning within the Horizon Europe project ARGENTIC. His work covers conversational AI, retrieval augmented generation, AI agents and human-AI interaction. He built the AI platform behind NeuroClima, an EU climate policy project, and is responsible for development in AIKAA, which brings generative AI to small Finnish companies. His first-author paper on AI-assisted spoken reviews appeared at ACM MUM 2024. ## Publications - [From Voice to Value: Leveraging AI to Enhance Spoken Online Reviews on the Go](https://doi.org/10.1145/3701571.3701593): ACM International Conference on Mobile and Ubiquitous Multimedia (MUM 2024), 2024, first author. Speak a restaurant review and let an AI help turn it into one you would post. A field study with 14 people and 157 reviews: willingness to share rose 31%, and confidence in writing a good review rose 88%. - [Factors of Trust for Successfully Adopting Social Robots on the Campus](https://doi.org/10.1145/3764687.3764717): Australian Conference on Human-Computer Interaction (OzCHI 2025), 2025, co-author. What students and staff would need before trusting a social robot on campus, from 59 people who met a Pepper robot first. Knowing what the robot does, and who is behind it, came first. - [A modular transformer framework for automated detection and structuring of social tipping point evidence in climate-related documents](https://arxiv.org/pdf/2609.12254): In preparation, first author. Finding the moments a public conversation turns, and giving that evidence enough structure that something else can act on it. The classifier behind it also runs inside NeuroClima. - What Does a Faithfulness Score Summarise? Evidence from a Production RAG System Built with Open-Source Technologies: In preparation, first author. Reading the explanations behind an LLM judge’s faithfulness scores for 1,197 answers from NeuroClima, and finding what a single number leaves out: about one in seven pieces of evidence the judge relied on was text a model had written, not the source. - AIRA Scholar-KG: A Multi-Source Knowledge Graph and Grounded Conversational Assistant for Academic Research Exploration: In preparation, co-author. Asking questions of the research literature and getting answers grounded in a knowledge graph assembled from several sources, rather than a single index. - Log2Sign: Automatic signature generation using LLMs with Explainability and Interpretability: In preparation, co-author. Generating intrusion detection signatures from logs automatically with large language models, while keeping each signature explainable and interpretable. ## Projects - [NeuroClima](https://www.kavinduravishan.com/projects/neuroclima/): EU Horizon project. Over 10,000 climate and policy documents made answerable in four languages, at whichever reading level the person asking needs. - [ARGENTIC](https://www.kavinduravishan.com/projects/argentic/): PhD research. My PhD. Every AI workload has to run somewhere, and something has to decide where. This is about making that decision learn. - [AIKAA](https://www.kavinduravishan.com/projects/aikaa/): ESF+ project. Find out which parts of the week a small Finnish company keeps repeating, and build something that takes those parts off them. - [Voice to Value](https://www.kavinduravishan.com/projects/voice-to-value/): Master’s thesis. You talk about an experience. The model turns it into something you would actually post. My master's thesis, and a paper at ACM MUM 2024. - [Human-Robot Interaction](https://www.kavinduravishan.com/projects/human-robot-interaction/): Summer internship. A Pepper robot in the corridors of the University of Oulu, and a study of what people would need before they trusted one on campus. A summer internship. ## Education - Master's in Computer Science and Engineering, major in Applied Computing, University of Oulu, Finland, 2024 - Bachelor's in Engineering Science, major in Telecommunications, Riga Technical University, Latvia, 2022 - Electrical and Electronics Engineering, SLTC Research University, Sri Lanka ## Areas Conversational AI, AI agents, Reinforcement learning, Multi-agent systems, Natural language processing, Large language models, Retrieval augmented generation, Knowledge graphs, LLM evaluation, Human-AI interaction, Human-computer interaction, Human-robot interaction, Voice user interfaces, MLOps, Kubernetes. ## Notes - [Retrieval over ten thousand policy documents](https://www.kavinduravishan.com/notes/retrieval-over-10000-policy-documents/): Making a corpus of climate and policy PDFs answerable in four languages: ingestion, GraphRAG reasoning, cutting sixty seconds to ten, and evaluating whether it worked. ## Optional - [About](https://www.kavinduravishan.com/about/): facts, timeline, bios to reuse and photographs - [Publications](https://www.kavinduravishan.com/publications/): every paper, with DOIs - [Full text](https://www.kavinduravishan.com/llms-full.txt): every page as plain text - [Résumé](https://www.kavinduravishan.com/resume.json): the same facts in the JSON Resume format - [Sitemap](https://www.kavinduravishan.com/sitemap.xml) ## Availability Open to research collaboration and product engineering work. Contact: ravishankavindu [at] gmail [dot] com.