CURRICULUM VITAE

Dilini Sumanapala

Dilini Sumanapala

Dilini Sumanapala

Solutions architecture for LLM-powered applications, agentic workflows, conversational AI platforms, and production NLP systems.

Solutions architecture for LLM-powered applications, agentic workflows, conversational AI platforms, and production NLP systems.

Solutions architecture for LLM-powered applications, agentic workflows, conversational AI platforms, and production NLP systems.

Founder, Genverv Ltd.

London, UK

Solutions Architecture · AI Agents · RAG · Enterprise AI Enablement

Solutions Architecture · AI Agents · RAG

Skills

Skills

A practical architecture stack for designing, integrating, evaluating, and operating language-driven solutions.

AI Agents

RAG

LangChain

LangGraph

Rasa

Python

NLP

LlamaIndex

OpenAI Deployment

AWS

Haystack

Hugging Face

Experience

Experience

AI Solutions Architect and LLM Applications Consultant

AI Solutions Architect and LLM Applications Consultant

10.2023 – Present · London, UK

Lead Solutions Architect — Conversational AI & Agents

Designed reference architectures for enterprise LLM orchestration, tool integration, system boundaries, evaluation workflows, and maintainable assistant deployments.

AI Agent Architecture SME, LangChain / LangGraph

Architected production-ready AI agent pipelines in Python using LangChain and LangGraph, including tool design, control logic, OpenAI integration, safety guardrails, and deployment guidance for 14,000+ DataCamp subscribers.

RAG Solutions Architect — Patent Intelligence

Designed an LLM-powered patent chat solution for biotech use cases, combining document ingestion, AWS Textract, retrieval design, vector storage, evaluation, and application delivery across LangChain, LlamaIndex, OpenAI, Hugging Face, Streamlit, and Qdrant.

Independent ML and NLP Solutions Consultant

Independent ML and NLP Solutions Consultant

04.2019 – 10.2023 · London, UK

Applied NLP solution design

Designed and delivered a RoBERTa-based sentiment analysis pipeline across 60,000 tweets, translating model outputs into brand and marketing decision support.

Recommendation architecture

Developed a document similarity and recommendation pipeline using Gensim tf-idf, structuring content hierarchies for discovery across fiction volumes.

Information extraction systems

Built SpaCy-based entity extraction workflows for 9,000 European Parliamentary documents, turning unstructured text into searchable people, organisation, place, event, and date signals.

Decision-support simulator

Created a PyCaret and Streamlit machine-learning app to model videogame recommendation likelihood and support future-state user behaviour planning.

Education

Education

PhD, Psychology (Cognitive Neuroscience)

2013 – 2017 · Bangor University, Wales, UK

Investigated MRI brain patterns involved in observational learning using machine learning.

Postdoctoral Researcher (Neuroscience)

2017 – 2019 · Birkbeck / UCL, London, UK

Managed delivery of a videogame-based maths and science learning intervention across 90 primary schools in England involving 6,500 pupils.

Publications

Publications

Sumanapala, D., Walbrin, J., Kirsch, L., & Cross, E. (2018). Neurodevelopmental perspectives on dance learning: Insights from early adolescence and young adulthood. Progress in Brain Research, 237, 243–277.

Kirsch, L., Diersch, N., Sumanapala, D., & Cross, E. (2018). Dance training shapes action perception and its neural implementation within the young and older adult brain. Neural Plasticity.