Denis

Denis Rizun
Stuttgart, Germany | [відкрити контакти](див. вище в блоці «контактна інформація») | LinkedIn: Denis Rizun | Telegram: @drizun
Summary
AI Engineer building production LLM systems end to end. Goes below the API layer when it pays off: fine-tuning,
embeddings, model evaluation. Backend and cloud engineering background. Comfortable owning technical
decisions and bringing AI practices into everyday engineering work.

Experience
AI Software Engineer, Go Wombat – Remote, Estonia May 2025 - Present
• Architected a monorepo agent platform serving 10+ independently deployed agents, cutting new-agent delivery
from 4 days to under 6 hours by extracting a shared orchestrator, tool registry, and context layer.
• Optimized a geospatial media-planning map handling 100M+ advertising screens to <500ms p95 viewport
response time through backend-driven spatial clustering, viewport-bounded queries, and zoom-aware data
delivery.
• Improved answer quality in a multi-tenant RAG system from 41% to 13% irrelevant responses through chunk
retrieval optimization, tenant-aware filtering, context construction, and answer validation.
• Prevented a migration to a 4x more expensive model by building an LLM-as-a-Judge eval suite with regression
scoring and side-by-side production scenario comparison, holding projected monthly inference costs flat.
Software Engineer, Neptunes – Remote, Ukraine May 2024 - May 2025
• Designed and deployed a standalone microservice for the Telegram ecosystem, which automated in-application
asset distribution and reduced manual workload by 90%, supporting 30K+ daily transfers.
• Optimized high-traffic database queries via partitioning, introducing caching strategies, reducing query response
time from 2s to 0.2s on datasets exceeding 4M+ rows.
• Migrated a critical business-data webhook service to a cloud-native, redesigning its processing pipeline to handle
high-throughput events reliably, which improved availability SLAs to 99.9%.
Software Engineer, Infinitum – Remote, Estonia Sep 2023 - May 2024
• Developed a centralized service for processing logs, transactions, and events, delivering real-time metrics for
analytics, improving observability and enabling analysis of user behavior patterns, while ensuring scalability to
50 RPS without performance degradation and reducing manual reporting time by 70%.
• Integrated a payment and subscription workflow using Stripe with recurring billing, webhooks, and refund
automation, which supported 10K+ monthly transactions.
• Refactored legacy test suites and added new automated tests, which increased coverage of critical business logic
from 65% to 90% and reduced production bugs by 35%.

Education
University of Stuttgart, BS in Computer Science Sept 2024 – Jul 2027
• Coursework: Application and implementation of AI in production and management

Technologies
Programming Languages: Python, Go, SQL
AI & LLM: RAG (chunking, hybrid retrieval, reranking), embeddings, semantic/vector search, agents (tool calling,
ReAct, multi-agent), LLM-as-a-Judge evals, LLM observability, prompt versioning, cost/token optimization
Models & Tooling: OpenAI (GPT, embeddings, Whisper), Anthropic Claude, Gemini, AWS Bedrock, Llama,
Hugging Face, LangChain, LangGraph, LangSmith, Langfuse, Pydantic AI, PyTorch, Sentence-Transformers
Frameworks & Libraries: FastAPI, Django(DRF, Ninja), SQLAlchemy, Alembic, Gin, GORM
Messaging & Async Processing: Kafka, RabbitMQ, Celery, TaskIq, Redis Streams / PubSub
Databases & Storage: PostgreSQL(inc. pgvector), MySQL, Qdrant, Weaviate, Pinecone, ClickHouse, Redis, S3
Cloud & DevOps: Docker, Kubernetes, AWS, GCP, Terraform, CI/CD (GitHub Actions), Nginx, Caddy
Languages: English (B2), German (C1), Ukrainian (Native), Russian (Native), French(A2)