01/profile.sh
$ whoami
Dmytro Yaremchuk — full-stack developer, AI-integrated
Full-stack Developer with 4+ years of experience building AI products — Front-end interfaces and Back-end APIs. Built CRM systems, AI assistants, and transaction flows handling 10k+ daily operations. Specialized in high-load, real-time interfaces.
- Years exp
- 4+
- Daily tx handled
- 10k+
- LCP improvement
- 2.3x
Ivano-Frankivsk, UkraineIvano-Frankivsk National Technical University of Oil and GasEnglish (B2) · Ukrainian (C2) · Russian (C2) · Polish (A2)
02/systems
Wahoo Predict
Next.js · TypeScript · WebSockets
- task
- Real-time prediction market — sports, esports, and 5-minute crypto price calls (BTC, ETH, SOL) — needed a UI that stays readable while odds and volume update continuously.
- constraint
- Market cards, countdown timers, and live odds/volume indicators had to keep updating in place under fast data flow without re-render jank or misreads under pressure.
- decision
- Optimized the market UI for live-updating state, and built the core app logic for odds recalculation, market resolution, and leaderboard ranking by PnL and prediction accuracy.
- result
- Running in production at wahoopredict.com, with live markets, leaderboards, and user profiles.
Spefix
Next.js · TypeScript · FastAPI · PostgreSQL
- task
- Build a search layer that returns semantically relevant results across large content sets, not just keyword matches.
- constraint
- Had to ship as a white-label component other client sites could embed, with a latency budget that holds under real-time query load.
- decision
- NLP-based vector search behind scalable FastAPI + PostgreSQL endpoints, fronted by a Next.js site built for SEO and CMS-friendly editing.
- result
- Running in production at spefix.com and integrated into client sites as a white-label search product.
Ailunara
React · TypeScript · Tailwind · LLM API
- task
- Let non-technical users create and manage their own AI assistants without touching model configuration directly.
- constraint
- LLM wiring (ChatGPT + Claude APIs, external service calls) had to sit behind a UI simple enough for the end user, while staying extensible for per-assistant personalization.
- decision
- Built the assistant builder in React/TypeScript on top of an LLM integration layer, plus an admin dashboard for indexed content, analytics, and fine-tuning model behavior with the AI team.
- result
- Live at lunara.bot — users configure and run their own assistants without writing integration code.
02/systems/dizair.tcdr
shipped
Dizair Software — CRM
React · WebSockets · Node.js
- task
- Enterprise CRM needed live schedule and workflow updates, not periodic polling, without degrading page performance.
- constraint
- The existing high-traffic dashboard page was loading in 4.2s, and the transaction UI had to hold up at 10k+ operations a day.
- decision
- Rebuilt the dashboard's data layer on WebSockets for live updates and reworked the render path for the transaction processing UI.
- result
- LCP dropped from 4.2s to 1.8s; the transaction UI now handles 10k+ operations/day in production.
03/stack
03/stack.manifest
- front-end
- React.js, Next.js, TypeScript, JavaScript ES6+, Tailwind, HeadlessUI, WebSockets, SSE
- back-end
- Node.js, Express.js, REST API, JWT, OAuth, FastAPI
- databases
- MongoDB, PostgreSQL, MySQL
- devops
- Docker, CI/CD, Nginx, Linux, Vercel, GitLab, Sentry
- testing
- Vitest, React Testing Library, Unit, E2E, Manual QA
- ai tools
- Claude API, ChatGPT API, Cursor, GitHub Copilot, LangChain
- ba analyst
- SQL, Power BI, Tableau, Excel, Power Query
04/contact
04/contact.sh
Open to full-stack and AI-integration work, remote or Ivano-Frankivsk.