Resume

Built an LLM-based legal-assistant on top of a semantic-search tool with:

• Product design and implementation
• Constraining AI agent via MCP tools to produce correct legal citations, reducing hallucination
• Built a TypeScript backend that uses the MCP (Model Context Protocol) to expose the legal database of Slovak legislation and external APIs as tools for the model
• Created the frontend chat interface with project and session management (TypeScript/React/Node)
• Implemented SSE (Server-Sent Events) to handle streaming and manage state with multiple LLM providers in a unified API.

Learning Rust, embedded/FPGA, basic AI/ML, and improving/exploring my language design/implementation skills:

• Rust DB server (team): Implemented the SQL interpreter and a persistence engine with custom binary format for tables and a GC to manage deletions
• Rust MLP (solo): Built a deep-learning MLP trained on MNIST
• Polymede Language/Compiler (Rust, WASM, solo)
• Thesis project: Designed a language based on Linear Type Theory and implemented interpreter for it in Rust.

Developed and maintained features in a distributed travel-tech platform, from Scala/PostgreSQL/GraphQL backend to Elm/TS Frontend:

• Onboarded senior backend developers into the Elm frontend
• Developed end-to-end features - from PostgreSQL schema design to responsive Elm UI for the customer facing website, including the implementation of i18n/localization pipeline, and analytics
• Delivered backend features in Scala in a distributed architecture, managing concurrency requirements and distributed cache
• Reduced payload size on the frontend. Reverse-engineered an undocumented third-party API to replace a bloated React embed with a lightweight JS/Elm, and translated UX designs into optimized, responsive frontend components (like lazy-loading image carousel in Elm).

• Quickly onboarded into a ~100K line Elm enterprise without prior Elm experience, implementing frontend features and UI components based on high-level architectural specifications
• Mentored colleagues who were new to functional programming, introducing them to Elm architecture, state-management to help them contribute to the codebase
• Built custom Elm parsers (including a CSV parser and a dataset query DSL) to overcome the limitations of existing libraries, focusing on clear error messages for end-users
• Maintained strict technical standards during code review. Argued against anti-patterns, actively engaged in constructive architectural debates to ensure highly performant, maintainable codebase.

Built small client websites and interactive frontend applications (React)

Implementing live, programmable knowledge environment (inspired by TiddlyWiki/Smalltalk) and setting up a self-hosted home-lab/web-server:

• Implemented a push-based Signals runtime in TypeScript, using Directed Acyclic Graph (DAG) with topological sorting for glitch-free value propagation
• Implemented transactional hot-swapping at runtime of live signal nodes with cycle-detection and automated rollback
• Deployed a self-hosted homelab on bare-metal NixOS, manually configuring Nginx reverse proxies, WireGuard VPNs, automated DNS/domain routing, and a personal Forgejo Git instance
• Visually designed and built my personal webpage to force myself to think more like a designer. Created my own consistent design system.