# Tomas Jankauskas

> AI Developer and Data Scientist. Data Science master's student at Columbia University (New York) and former AI Developer at Hostinger International, where he pioneered company-wide adoption of the Model Context Protocol (MCP) and built LLM-powered products. Background: senior frontend engineering (Vue 3 / Nuxt / React) and applied physics. Open to AI, ML, and frontend roles.

## About

- Currently: M.S. in Data Science, Columbia University, New York (Oct 2025 - present).
- Previously: AI Developer at Hostinger International (one of Europe's fastest-growing companies), then ~4 years there as a Senior Frontend Developer (Vue 3 / Nuxt) shipping to 4.6M+ users.
- Started as a frontend engineer and trained as a physicist (B.S. Applied Physics, Vilnius University).
- Based in New York, NY. From Lithuania. Has lived in Shanghai, South Korea, and the US. 26 countries visited.

## Focus areas

- **AI / ML:** LLM APIs, Model Context Protocol (MCP), RAG, prompt engineering, PyTorch, vector search, structured outputs with Pydantic.
- **Frontend:** Vue 3, Nuxt, React, React Native, TypeScript.
- **Backend & data:** Python, FastAPI, Node.js, Scrapy, SQL/Mongo.
- **Cloud & tooling:** AWS, Docker, GitHub Actions, Grafana, Amplitude.

## Selected work

- Pioneered company-wide adoption of the Model Context Protocol (MCP) at Hostinger - helped automate and deflect ~68% of routine support interactions for 4.6M global users, contributing to ~€9M in annual operational savings.
- Built custom MCP servers integrating Amplitude telemetry with live UI adjustments via Claude Desktop for multi-locale A/B text experimentation, and enforced structured JSON outputs via strict Pydantic schemas (formatting exceptions cut from 28% to <0.5%).
- At a company hackathon, our team won 2nd place integrating MCP into WordPress to build features for users.
- Led frontend for the Hostinger Pro Panel (Agency Hosting) launch - centralized multi-site dashboards and client collaboration tools.
- Engineered an AST-based Python refactoring script that migrated 1,000+ hardcoded frontend files into i18n translation slugs in under 3 minutes (saving 80+ hours of manual work).
- Designed hPanel customer onboarding in Vue 3, driving a ~20% relative lift in registration conversion across 4.6M active clients.

## Links

- Website: https://tomasjankauskas.com/
- LinkedIn: https://www.linkedin.com/in/tomasjank/
- GitHub: https://github.com/TomasJank
- Email: tomastomasj@gmail.com
- Full details: https://tomasjankauskas.com/llms-full.txt