Full-stack engineer. I ship production web apps end-to-end, then put AI agents to work inside them.
TypeScript and C# across the stack: Next.js and React on the front, ASP.NET Core and Node (Express, and NestJS in progress) on the back, with a growing amount of LangGraph, RAG, and n8n wiring in between. Computer Science @ Tbilisi State University · building real client software from Tbilisi, Georgia 🇬🇪
I'm a Computer Science student who spends most of his time shipping software that other people actually use: client sites, hackathon products, and internal tools, not just coursework. I work across the whole stack: designing REST APIs and data models, building interfaces in React and Next.js (which I also lean on as a full-stack framework), and keeping systems clean enough to change later. I care about clean architecture on whichever side of the stack I'm working. Lately I've been pulling AI agents into that work, from a LangGraph travel planner with human-in-the-loop control to n8n automation pipelines behind a rental platform.
I like problems where product thinking and engineering meet: figuring out what to build, then building it in a way I won't regret in six months. Currently in the HIGGS AI Talent Club, deepening applied ML and LLM engineering.
- 🔭 Shipping full-stack apps for real clients, plus a three-sided marketplace and a mood-tracking app
- 🌱 Going deeper on AI agents, RAG systems, and clean backend architecture at scale
- 🗣️ Georgian (native) · English (professional) · Russian (professional)
- 💬 Ask me about ASP.NET Core, Next.js, agentic workflows, or how to design an API you can live with
Finding student housing in Georgia usually means scrolling Facebook groups and gambling on a stranger. SakhliAI scores compatibility on what actually causes flatmate conflict (sleep schedule, cleanliness, budget, lifestyle), then hands the tedious parts (bill-splitting, house rules, lease prep, channel sync) to an AI agent. For landlords, it's a control center that runs hybrid short-/long-term rentals on autopilot.
• Stack: React 19 · TanStack Start (SSR) · Tailwind v4 · shadcn/ui · Supabase (Auth, Postgres + RLS, Realtime) · n8n agentic workflows · bilingual EN/KA • Why it's interesting: the automation is a visible character in the UI. A live agent-activity feed surfaces every decision the system makes. Built at the Nakadi Hub Vibe Coding Hackathon. • Links: Repo · Live demo
An AI agent that plans real trips end-to-end and knows when to ask a human before committing. Built on a cyclical LangGraph state machine that reasons over requirements, calls real-world tools for live data, and pauses for approval before finalizing an itinerary.
• Stack: Python · LangGraph · Google Gemini 2.5 Flash · human-in-the-loop interrupts · MemorySaver for cross-thread state · tool integrations (DuckDuckGo, Wikipedia, Open-Meteo, Nager.Date)
• Why it's interesting: it's a genuine agent architecture (reason → act → verify) with real safety rails, not a single prompt. Built during the HIGGS AI Agents course.
• Links: Repo · Video demo
A full ASP.NET Core Web API connecting artists and critics: portfolios, artwork uploads, professional criticism, follows, and role-based admin. The project I point to for backend depth on a real, multi-role service.
• Stack: ASP.NET Core · Entity Framework Core · JWT auth · Hangfire background jobs · AWS RDS + S3 · Repository pattern · AutoMapper · FluentValidation · pagination • Why it's interesting: it's structured the way a maintainable service should be, with clear layering, validation at the edges, background processing, and cloud storage, rather than everything crammed into controllers. • Links: Repo
Built for the NASA Space Apps Challenge 2025. Two halves: a Next.js app for astronaut-style health assessment and space-biology education, and a fully offline RAG assistant grounded in 33+ space-medicine research papers running a local LLM.
• Stack: Next.js (frontend) · Python RAG pipeline · vector database · Ollama (local LLM) · authentication + health questionnaire flow • Why it's interesting: the RAG side runs entirely offline. A deliberate constraint that forced real retrieval engineering instead of leaning on a hosted API. • Links: Repo
A production portfolio and blog platform for a practicing surgeon, with a role-based CMS, draft/publish workflows, live preview, and one-click AI localization across Georgian, English, Hebrew, and Arabic. Optimized with static generation, lazy loading, and CDN asset delivery.
• Stack: Next.js · Payload CMS · PostgreSQL · Cloudflare R2 · Framer Motion · Vercel • Links: Live site (private repo, client project)
Only what I actually build with.
| Category | Technologies |
|---|---|
| Languages | |
| Frontend | |
| Backend | |
| Data | |
| AI / Agents | |
| Cloud & DevOps | |
| CMS & Tools |
| Role | Where | When |
|---|---|---|
| Full-Stack Engineer | Freelance / Client Work | Feb 2026 - Present |
| AI Research Trainee | HIGGS AI Talent Club | Nov 2025 - Present |
| Front-End Developer | Sivrce AI | Jan 2026 - Apr 2026 |
| Junior Front-End Developer | ROWIX | Dec 2025 - Apr 2026 |
| Full-Stack Developer | Koober Coders | Jun 2025 - Sep 2025 |
What I actually did →
- Freelance / Client Work: Design, build, and deploy web applications for real clients end-to-end: responsive frontends, backend APIs, database integration, and direct client collaboration on requirements. Shipped a multilingual medical CMS for a practicing surgeon (Next.js + Payload CMS + Cloudflare R2) with one-click AI localization across four languages.
- HIGGS AI Talent Club: Selected for a competitive applied-ML program. Built and evaluated n-gram / bigram language models, text-preprocessing pipelines, and NLP workflows; completed GPT / prompt-engineering training with the TSU Knowledge Transfer & Innovation Center and collaborated on AI prototypes with engineers and mentors.
- Sivrce AI: Built core frontend for an AI-powered recommendation platform with React, Next.js, and Strapi; reusable component systems and REST integrations powering personalized content delivery.
- ROWIX: Shipped production frontend features for client-facing products in React + TypeScript; reusable UI systems, responsive interfaces, REST integration, and state management, with code reviews in Agile cycles.
- Koober Coders: Full-stack delivery with Angular, ASP.NET Core, EF Core, and SQL Server; designed relational schemas and RESTful APIs supporting real business workflows.
Certiport IT Specialist - Software Development · HIGGS AI Neural Networks, Gen AI, n8n Automation and AI Agents · Volunteering at Global Tech Weekend Tbilisi 2026 · NASA Space Apps Challenge 2025 · Harvard CS50 - Python · DataArt JavaScript Workshop · Learn Swift (Codecademy)


