Senior Frontend Engineer
Amazon Web Services
Verify on Credly: AWS Solutions Architect – Professional (opens in a new tab)Amazon Web Services
Verify on Credly: AWS Developer – Associate (opens in a new tab)MongoDB
Verify on Credly: MongoDB Associate Developer (opens in a new tab)Production systems I designed and led — the decisions, the trade-offs, and what I’d do differently.
Bank of Georgia public site
Added an LLM layer in front of the existing keyword index so questions and typos resolve to indexed terms, plus a summary card built from the most relevant result pages. Model calls fall back to the raw query, so search keeps working through a provider outage.
~50%more clicks on search results
Bank of Georgia internal multi-locale content management system
Migrated a daily-use editorial CMS off Polymer 3, running old and new apps side by side so editors could keep working. Most of the work was rewriting the form engine that generates the whole editing UI at runtime from user-defined content types, including multi-locale editing.
5product teams migrated
High-traffic Bank of Georgia public site
Split a web-component SPA serving two product lines into a shell and two independently deployed remotes that share one copy of the framework runtime. Routes moved to their owning remote one content type at a time.
~10releases per month, up from ~6
High-traffic Bank of Georgia public site
Moved the design system and framework runtime out of three application bundles and onto a CDN, resolved by the browser from one import map. A design-system upgrade became a version bump in one JSON file.
~60%smaller per-application bundles
Bank of Georgia public sites · site search
Built bilingual full-text search and autocomplete for three public sites as one Node service with a hand-built character trie. Prefix lookup matches inflected Georgian nouns, and query-time transliteration lets Georgian typed in Latin letters find the same pages.
3public sites with bilingual search
High-traffic Bank of Georgia public site
Built an Express middleware that gives search crawlers finished, script-free HTML from the same origin that serves people. The key decision was forcing the shadow-DOM polyfill so content built from custom elements serializes.
5public sites with improved search indexing
Things I built on my own time to dig into a specific problem.
Focus: Data Integrity & AI Integration
Job application tracker built with Next.js 16, React 19, PostgreSQL, Drizzle ORM, and shadcn/ui. Users sign in with GitHub, track applications from wishlist to offer, and fill in new ones from a pasted job posting with Claude. Includes keyset pagination, optimistic updates, and race-safe daily limits on AI usage.
Focus: LLM Integration & Prompt Engineering
Front-end diagram generator built using React, TypeScript, and Vite, powered by Anthropic's Claude Haiku model and rendered client-side with Mermaid.js. The app turns a plain-English description of a flow (OAuth login with refresh tokens) into valid Mermaid syntax and renders it instantly as an SVG diagram in the browser — with copy-to-source and SVG export, and no backend required.
Focus: High-Concurrency & Real-Time Systems
A fee-aware, real-time arbitrage engine architected for low-latency financial data. I implemented a multi-threaded architecture using Web Workers to decouple high-frequency WebSocket streams from the UI thread, maintaining a locked 60 FPS. The project utilizes a strict TDD workflow with over 50 tests to ensure mathematical accuracy in fee-adjusted profit calculations across multiple exchanges.
Focus: Framework Internals & Performance
Excel application created with vanilla JavaScript. Implemented custom web framework from the scratch (with state, routing, eventing, storage, DOM manipulations). Used Tailwind CSS for styling.
Focus: Algorithmic Efficiency
JavaScript search engine built on a Trie (prefix tree) data structure for fast text lookup and autocomplete. Supports instant prefix-based searching, smart suggestions, and relevance-ranked results.
