Mantas Šimonis – AI Engineer & AI Technologies Specialist

I design and ship AI that improves real workflows – not slideware.

Hands-on across the whole AI lifecycle – idea validation, architecture, deployment, and the continuous improvement nobody puts on the slides. I've built production AI apps, advised colleagues on strategy, contributed to international initiatives, and helped organisations find where AI actually creates measurable value.

My world is large language models, RAG, AI-powered automation and Python software engineering – the engineering that takes a clever prototype and makes it dependable. The web fluency comes from a bachelor's in Web Page, Digital/Multimedia & Information Resources Design, before the master's in AI.

When I'm not in a terminal, I'm usually under the hood of a car, watering one of 30-odd houseplants, or somewhere abroad – 18 countries and counting. Same instinct either way: take something that works, and make it work better. More on the life page →

CASE_01 · internal tool · RCL

AI Expert Selector

Owned solo, idea → deployment. Reads a proposal, ranks the best-fit experts across the Council's whole database, and clusters projects by semantic theme.

OllamaEmbeddingsPythonSemantic clustering

PROBLEM

Matching each proposal to a specialist by hand took ~45 min – 75+ hours per funding call across 100+ proposals – and experts could be missed or over-assigned.

APPROACH

Embeds every proposal and all 2,000+ expert profiles into the same vector space (method from a published paper), ranking by semantic fit with the reasoning kept visible.

BUILT

A fully local Ollama pipeline – no proposal or expert data ever leaves the Council's servers – that load-balances so the same expert isn't repeatedly over-assigned.

OUTCOME

Selection now takes seconds per proposal, ~95% faster, fully explainable – adopted as an internal tool across the organisation.

45min → sec

per proposal, manual → AI

~95%

less time spent selecting

2,000+

experts matched, 100% local

CASE_02 · RAG chatbot · personal project

HR Knowledge Assistant

A personal project – a RAG chatbot for an HR use case: ask in plain language, get answers pulled from internal documents and colleague profiles – instead of digging through folders.

→ outcome: self-serve answers, less repeat Q&A for HR

RAGVector searchLLM

CASE_03 · AI due-diligence · RCL

AI Expert Analysis

Researches an expert with AI and surfaces what matters – automatically flagging conflicts of interest (COI) and other checks that used to be done by hand.

→ outcome: faster, auditable vetting

LLMCOI detectionPython

CASE_04 · gesture control · personal project

Medical Mouse

A personal project – lets doctors control a computer with hand gestures, so they never have to touch a mouse or keyboard (and take off sterile gloves) mid-procedure.

→ outcome: gloves stay on, workflow stays sterile

Computer VisionGesture ControlPython

AI that earns its place in the workflow.

I design and ship AI that improves real workflows – not slideware.

Selected projects.

AI Expert Selector

HR Knowledge Assistant

AI Expert Analysis

Medical Mouse

The stack, embedded.

Work with me.

Agent development

AI automation

AI consultancy

Part-time AI Officer