AI Engineer currently focused on voice AI and generative AI systems — building real-time multilingual voice agents (Uzbek, Russian, English), agentic LLM assistants that call tools to complete real-world tasks, RAG pipelines on OpenAI embeddings and pgvector, and workflow automation with n8n.
Shipped Uzbek voice pipelines handling the hard parts of low-resource-language speech (latency, turn-taking, phonetic normalization), not just API integration.
Brings a strong data science and security engineering background — production fraud-detection, device-fingerprinting, and TLS-fingerprinting systems — that grounds the AI work in engineering built to hold up under real traffic.
Hover / tap a skill for proficiency
01 / 08 · 2026
Turned Chatwoot's open-source core into a working AI customer-support product for the Uzbekistan market — rebranded, fully localized into Uzbek, and extended with a multi-tenant AI agent layer: RAG over each client's documents, a temporal-memory knowledge graph, and human handoff.
View project →
02 / 08 · 2025
An AI front desk where one agent handles customers across chat, phone, and automation — completing orders, forms, and API actions mid-conversation in Uzbek, Russian, and English. I built the production voice stack, the chat widget, and the automation layer.
View project →
03 / 08 · 2026
An agentic pediatric voice assistant that turns a 2 a.m. question into a clear next step, grounded in the child's own medical profile — speech-to-speech in Uzbek, Russian, and English via the OpenAI Realtime API.
View project →
04 / 08 · 2025
An enterprise LLM-visibility platform that ranks which websites most influence how a brand appears in AI outputs. I led the team and built the scoring agent and the metrics pipeline.
View project →
05 / 08 · 2025
Production research to cut RAG and LLM API costs through semantic caching — safely. Turned a headline "55% overlap" figure into a defensible ~10–15% cacheable scope with strict query gates.
View project →
06 / 08 · 2025
A production device-fingerprinting system for fraud prevention that re-identifies devices even as OS, fonts, timezone, or IP change over time — core engine in Go with Bayesian probabilistic matching.
View project →
07 / 08 · 2024
An internal red-team framework that hardens a security company's own bot-detection products with realistic, labeled adversarial data. Became the primary tool for benchmarking detection against the broader market.
View project →
08 / 08 · 2024
Probed the detection layers of a decentralized, blockchain-based security subnet with realistic behavioral-simulation models — custom kinematic and statistical models that scored a perfect 1.0 across challenge versions.
View project →Also: sentiment & NER analysis on news headlines (Inha University, 2024).
B.B.A., International Business and Trade · GPA 4.1 / 4.5
Selected coursework
Algorithms and Data Structures · Object-Oriented Programming · Computer Security · Discrete Mathematics · Introductory Engineering Mathematics · International Finance and Banking
Member, GDGoC (Google Developers Group on Campus), Inha University · Mar–Jun 2025
Certifications & training
Languages