Vladyslav

VLADYSLAV ZELINSKYI
[open contact info](look above in the "contact info" section) • Linkedin
• [open contact info](look above in the "contact info" section) • Lviv, Ukraine
SUMMARY

Production-focused AI/ML Engineer with deep expertise in applied GenAI and LLM architectures (agentic
RAG, multi-provider model routing, structured outputs), expert-level Python/SQL, and strong product sense.
I have extensive experience in building full-lifecycle AI applications, reducing inference costs, and improving
model accuracy to solve complex business challenges.
EXPERIENCE

Reply.io Remote
AI Engineer / Backend Developer Jun 2026 – Present
• Significantly reduced LLM inference costs across core production pipelines by building a privacy-aware
routing engine that directs public data to open-weight models (strict JSON-mode, 7 to 10s fallback SLA)
while isolating sensitive client documents on enterprise providers.
• Substantially increased NLP intent classification precision while maintaining full recall on critical
compliance triggers by re-architecting text analysis into a multi-label system, eliminating false positives
on ambiguous unstructured inputs.
• Minimized LLM hallucinations and accelerated production incident resolution by deploying a 6-tier
automated quality audit rubric, grounding outputs in verified domain data, and shipping JSON-schema
fixes based on deep log and event-stream diagnostics.
Xedrum Remote
Python AI Developer Feb 2026 – Jun 2026
• Fully responsible for designing and implementing a multi-stage AI extraction pipeline converting
unstructured documents into Markdown and RAG-ready records via Pydantic validation, taxonomy
controls, and enum normalization.
• Significantly increased ingestion throughput and reduced latency by restructuring background extraction
flows with async Python, batching, concurrency tuning, and adaptive rate-limit control.
• Automated end-to-end data delivery and improved LLM cost visibility by building REST APIs integrated
with S3/MinIO and Airtable, backed by structured logging, token usage tracking, and execution
diagnostics.
SoftServe Remote
Gen AI Engineer 2025 – Apr 2026
• Fully responsible for architecting a multi-service enterprise AI platform unifying autonomous ReAct agent
orchestration (LangGraph, AWS Bedrock) with an MCP-based RAG integration for dynamic tool
discovery and graph-expanded semantic retrieval.
• Streamlined real-time inventory and device monitoring by developing asynchronous FastAPI
microservices, shared typed-contract packages, and smart-query data access layers using PostgreSQL and
SQLAlchemy.
• Minimized production regression risks and accelerated debugging by establishing LangSmith-compatible
LLM tracing, structured logging, and automated integration testing suites using pytest and Testcontainers.
SKILLS

Technical: Python (FastAPI, AsyncIO, Pydantic, SQLAlchemy, NumPy, Pandas, SciPy), LangGraph,
LangChain, AutoGen, MCP, Agentic RAG, Semantic Search, PostgreSQL (pgvector), Redis, SQL, AWS
(Bedrock, Lambda, S3), Azure (OpenAI, App Service, Blob Storage), GCP Vision API, Docker, Git.
AI-Augmented Dev & Tools: Cursor, Claude Code, GitHub Copilot, LangSmith, pytest, Testcontainers, Jira,
Confluence.
Languages: English (Upper-Intermediate / B2), Ukrainian (Native).
EDUCATION

Kharkiv National University of Radio Electronics Bachelor's degree in Artificial Intelligence