Личные данные скрыты

IHOR LEVCHENKO
Full Stack TypeScript Engineer | AI / LLM Integration
Ukraine | Remote | English B2 | Ukrainian & Russian - Native

PROFESSIONAL SUMMARY
Full Stack TypeScript Engineer with 10+ years of experience building scalable web, real-time, payment, media, and
communication systems. Strong backend expertise in NestJS, Node.js, MongoDB/Mongoose, PostgreSQL, Redis, Stripe,
WebSockets, and Docker, with frontend experience across Angular, Next.js, React, Vue, and Nuxt. Practical experience integrating
OpenAI-based conversational and voice solutions, including prompt design, structured AI workflows, speech processing, and IP
telephony. Designed a production-oriented RAG architecture using embeddings and Qdrant for knowledge retrieval, while keeping
financial and identity decisions deterministic in backend services.

CORE COMPETENCIES
OpenAI SDK, conversational agents, prompt engineering,
AI / LLM structured outputs, Realtime Voice, Whisper/STT, TTS, RAG
architecture, embeddings, Qdrant/vector search
TypeScript, Node.js, NestJS, Express.js, REST APIs, WebSockets,
Backend
Socket.IO, BullMQ, Redis, Stripe, JWT, RBAC
Angular, Next.js, React, Redux Toolkit, Vue 3, Nuxt 3, Pinia,
Frontend
responsive UI, SSR
MongoDB, Mongoose, PostgreSQL, Prisma, Supabase, MySQL,
Data & DevOps Docker Compose, Nginx, CI/CD, Grafana, Loki, Prometheus
architecture

PROFESSIONAL EXPERIENCE
Full Stack / AI Integration Engineer - BurningHearts.online | 2024 - Present
● Develop and modernize a multi-application platform built with Angular, NestJS, TypeScript, MongoDB/Mongoose, Redis, Stripe,
Docker, Nginx, and shared TypeScript contracts.
● Restored and hardened Stripe checkout, webhook verification, payment history, and coin-wallet flows, including idempotent
credits, transaction-ledger records, atomic balance updates, and retry-safe MongoDB transaction handling.
● Implemented and debugged Angular SSR flows, browser/server safety, dependency injection, Docker runtime dependencies, and
shared-contract package integration across the workspace.
● Designed AI-assisted support scenarios that combine deterministic MongoDB/Stripe diagnostics with LLM-generated
explanations, ensuring the model does not control balances, payments, or authorization decisions.
● Designed a RAG-oriented knowledge workflow using embeddings and Qdrant: policy documents are chunked, vectorized, filtered
by metadata, retrieved semantically, and supplied to an LLM together with verified backend facts.
● Designed an AI-assisted verification-photo workflow in which a vision-capable model extracts observable evidence and a
handwritten visual code, while backend services perform deterministic code matching and route uncertain cases to manual
review.
● Integrated observability patterns with Grafana and Loki, and designed Prometheus-based metrics for API, Qdrant, RAG, LLM
latency, fallback rates, and manual-review flows.
● Used AI-assisted development tools including Cursor, Windsurf, and coding agents for code analysis, migration planning,
debugging, and controlled implementation workflows.
Technologies: Angular, NestJS, TypeScript, MongoDB, Mongoose, Redis, Stripe, Docker Compose, Nginx, OpenAI SDK, Qdrant
architecture, Grafana, Loki, Prometheus architecture, WebSockets, shared-contracts
Full Stack Engineer - Real-Time Transportation Platform | 2024
● Built NestJS backend services for real-time ride matching, trip scheduling, driver availability, and asynchronous task execution
using BullMQ and Redis.
● Integrated Prisma with Supabase/PostgreSQL, applying transactional and ACID principles to user and trip operations.
● Developed real-time driver/passenger communication with Socket.IO and responsive Next.js/React interfaces with Redux Toolkit.
● Containerized the application with Docker and optimized frequently accessed geolocation and operational data through Redis
caching.
Technologies: NestJS, Next.js, React, TypeScript, Prisma, Supabase, PostgreSQL, Redis, BullMQ, Socket.IO, Docker, Google
Maps API
Full Stack Developer - Vue/Nuxt Modernization | 2023 - 2024
● Migrated a legacy Vue 2/Vuex/Vuetify 2 application to Vue 3, Nuxt 3, Pinia, Vuetify 3, Composition API, SSR, and Nitro server
routes.
● Implemented JWT authentication, role-based access control, Zod validation, Knex.js data access, migrations, filtering, and
structured error handling.

Ihor Levchenko - Full Stack TypeScript Engineer | AI / LLM Integration
 ● Configured Docker Compose environments for frontend, backend, and database services.
Technologies: Vue 3, Nuxt 3, Pinia, Vuetify 3, TypeScript, Knex.js, JWT, Zod, Docker Compose
Full Stack Developer - E-commerce & Web3 Platform | 2018 - 2023
● Developed a scalable e-commerce platform with Next.js, React, Redux Toolkit, MongoDB, secure payment flows, dynamic
catalog functionality, and responsive interfaces.
● Integrated Web3Modal and Ethereum wallet connectivity for multi-chain user flows.
● Implemented real-time chat functionality using NestJS and Socket.IO.
Technologies: Next.js, React, Redux Toolkit, NestJS, Prisma, MongoDB, Docker, Web3.js, Viem, TypeScript
Full Stack Developer - Real-Time Communication & Voice Systems | 2016 - 2018
● Built real-time chat, audio, and video communication using React, Express.js, Socket.IO, WebRTC, Redux, and MySQL.
● Designed and maintained Asterisk PBX, VoIP call routing, recording, multi-level IVR, and speech-recognition workflows.
● Integrated voice-recognition capabilities for automated customer interactions and later applied this telephony background to
OpenAI Realtime Voice-based agent scenarios.
Technologies: Node.js, Express.js, React, MySQL, Socket.IO, WebRTC, Redux, Asterisk PBX, IVR, VoIP, Speech Recognition APIs

AI / LLM EXPERIENCE HIGHLIGHTS
● Integrated OpenAI-powered conversational agent functionality for domain-specific customer interaction, including a
pharmaceutical-assistant use case.
● Worked with OpenAI Realtime Voice SDK concepts and IP-telephony integration for real-time voice interactions, building on
practical Asterisk/IVR experience.
● Implemented or designed backend-only provider integration patterns: API-key isolation, provider services, prompt assembly,
structured responses, timeout handling, and deterministic fallback templates.
● Designed RAG workflows in which verified business facts come from MongoDB/Mongoose and Stripe, while Qdrant retrieves
policy and support knowledge through embedding-based vector search.
● Defined safe AI boundaries: LLMs explain results but do not modify balances, approve identities, execute unauthorized tools, or
replace deterministic financial logic.

TECHNICAL SKILLS
Languages: TypeScript, JavaScript, SQL
Frontend: Angular, Next.js, React, Redux Toolkit, Vue 3, Nuxt 3, Pinia, Tailwind CSS, Material UI
Backend: Node.js, NestJS, Express.js, REST, WebSockets, Socket.IO, BullMQ, JWT, RBAC
AI / Voice: OpenAI SDK, prompt engineering, conversational agents, Realtime Voice, Whisper/STT, TTS, structured AI workflows,
RAG and vector-search architecture
Databases: MongoDB, Mongoose, PostgreSQL, Supabase, Prisma, MySQL, Redis
DevOps / Observability: Docker, Docker Compose, Nginx, CI/CD, Grafana, Loki, Prometheus architecture, AWS, Google Cloud
Other: Stripe, WebRTC, Asterisk PBX, IVR, VoIP, Web3, Google Maps API, ACID principles

LANGUAGES
English: B2 | Ukrainian: Native | Russian: Native

Ihor Levchenko - Full Stack TypeScript Engineer | AI / LLM Integration