AI agents & automation
Agents that do real work: handle requests, operate your CRM and ERP, call APIs and hand complex cases over to a human.
- Architecture: workflow or agent
- Tools and MCP servers
- Access control & human in the loop
Available for new projects
AI/ML Developer · AI Agents · RAG · LLM
AI that actually works
for your business
I'm Volodymyr Sydorenko, an AI/ML developer. I design and build AI agents, RAG assistants and LLM integrations — from a quick prototype to a reliable production system with measurable results.
Volodymyr Sydorenko
AI/ML Developer
Kyiv · working remotely
- AI agents
- RAG
- LLM integrations
- Evals
$agent.run("your_process")
✓ retrieved context · 4 sources
✓ tools called · crm.update, mail.send
✓ eval score · 0.94
- 3+
- AI agents in production
- 30+
- articles on AI engineering
- < 24h
- response time
Services
From idea to an AI system in production
I own the whole path: feasibility, architecture, development, integration with your systems and support. You talk directly to the engineer who writes the code.
RAG & knowledge assistants
An assistant that answers from your documents, knowledge base and data — with source citations instead of guesses.
- Indexing, chunking, embeddings
- Hybrid search and reranking
- pgvector, Qdrant, Pinecone
LLM integration into your product
I embed GPT, Claude or open-source models into your app: streaming chat, structured outputs, generation and classification.
- Structured outputs & tool calling
- Retries, fallbacks, rate limits
- Python, TypeScript, Java (Spring AI)
Document processing
Automatic extraction from invoices, contracts, forms and scans into structured data — with validation and human review.
- OCR + LLM extraction
- Schema validation
- Integration with accounting systems
Evals, monitoring & cost
I make quality measurable: test sets, LLM-as-judge, tracing and token cost control, so the system never degrades silently.
- Eval sets and regression tests
- Tracing and observability
- LLM cost optimization
Private & self-hosted models
When data can't leave your perimeter: open-source LLM deployment, task-specific fine-tuning, PII protection, GDPR and EU AI Act compliance.
- vLLM, Ollama, Hugging Face
- Fine-tuning vs RAG vs prompting
- PII filtering and security
Cases
What I've built
Real estate · Telegram
AI real estate sales agent in Telegram
The agent talks to clients in Telegram: it finds out their budget, preferred area and requirements, picks matching listings from the database and answers questions. Clients ready to move forward are handed over to a manager with the full conversation history.
24/7available to clients, no days off
Sales · Voice AI
Voice AI agent for outbound calls to prospects
A prototype agent that calls through a contact list on its own, holds a natural voice conversation, qualifies interest and logs the outcome of every call. Interested prospects are passed on to a manager.
Prototypeautomated calling and lead qualification
Why me
An engineer, not a middleman
01
Direct line to the developer
No account managers, no broken telephone. You talk to the person who designs and writes the code.
02
Production-first
No demos for the sake of demos: eval sets, monitoring, retries and cost control are built in from the start.
03
Security & privacy
Prompt-injection defenses, PII handling, GDPR and EU AI Act requirements. Models inside your perimeter when needed.
04
You own the code and knowledge
Full repository access, documentation and team training. No vendor lock-in.
Engagements
Pick where to start
Every project starts with a short free call. After it you get a work plan and an estimate broken down by phase.
1–2 weeksFixed price
AI audit & roadmap
Find out where AI will pay off and where it won't — before any development starts.
- Process and data analysis
- Prioritized use cases with ROI estimates
- Architecture and model selection
- Implementation plan and budget
2–6 weeksFixed scope
Prototype / MVP
A working prototype on your data to validate the idea before a bigger investment.
- Working agent or RAG assistant
- Eval set and quality metrics
- Integration with 1–2 systems
- Demo and recommendations report
1 month +Monthly
Build & support
Full development, production launch and ongoing evolution of your AI system.
- Production architecture and deployment
- Monitoring, tracing, cost control
- Regular releases and reports
- Support and quality improvements
How I work
A transparent process with no surprises
01
Discovery
A free call to break down the task: goals, data, constraints, success metrics. I'll tell you honestly if AI isn't needed.
02
Prototype
A fast prototype on real data plus an eval set, so quality is measured before you invest more.
03
Build
One- to two-week iterations with demos. Repository access from day one — you see every step.
04
Launch & Support
Deployment to your infrastructure or cloud, monitoring, documentation and knowledge transfer to your team.
Technology
The tools I work with
LLM & Models
- OpenAI GPT
- Anthropic Claude
- Llama · Mistral · Qwen
- Hugging Face
- Embeddings · Rerankers
Agents & RAG
- LangGraph
- Model Context Protocol
- Spring AI
- Tool calling
- Structured outputs
Data & Search
- PostgreSQL + pgvector
- Qdrant
- Pinecone
- Weaviate · Milvus
- Elasticsearch
ML & Serving
- Python
- vLLM
- Ollama
- Fine-tuning (LoRA)
- FastAPI
Evals & Observability
- Langfuse
- OpenTelemetry
- Promptfoo
- LLM-as-judge
- Cost tracking
Product & Infra
- TypeScript · Next.js
- Java · Spring Boot
- Docker
- Kubernetes
- GitHub Actions
FAQ
Frequently asked questions
Which models do you work with?
OpenAI, Anthropic Claude and open-source models (Llama, Mistral, Qwen). I pick the model for the task — by quality, cost, latency and privacy requirements — and keep it swappable.
Will my data be safe?
Yes. I sign an NDA, work with least-privilege access and can deploy the whole solution inside your infrastructure with self-hosted models, so data never leaves your perimeter.
How much does an AI project cost?
It depends on scope and integrations. The audit and prototype are fixed-price, so you know the cost before we start. After a free call I send an estimate broken down by phase.
How long does it take to launch?
A prototype on your data usually takes 2–6 weeks. A full production system takes one to three months, depending on the number of integrations and quality requirements.
Do you work with companies outside Ukraine?
Yes, I work remotely with clients in Ukraine, Europe and the US. Communication in English or Ukrainian, with calls in your time zone.
What if AI isn't a fit for my problem?
I'll tell you so on the first call. Sometimes simple automation or a classic algorithm gives a better result for less money.
Blog
Latest writing on AI
AI engineer roadmap 2026: skills a software developer needs to build LLM products
What an AI engineer actually does, how the role differs from ML engineering, and a…
How to choose an LLM for your product: benchmarks, real evals, cost, latency and lock-in
A practical framework for selecting language models: what public benchmarks like…
Privacy for LLM applications: GDPR, personal data in prompts, PII redaction and vendor choices
A developer's guide to personal data in AI features: GDPR principles applied to LLMs,…
Document processing with LLMs: extracting data from invoices, contracts and scans into your ERP
How to build an intelligent document processing pipeline with vision-capable LLMs: PDF…
Reliable LLM integrations: timeouts, retries, rate limits, fallbacks and graceful degradation
How to build LLM features that survive provider outages and rate limits: error types,…
Spring AI for Java and Kotlin teams: chat, RAG, tools and structured output in Spring Boot
A practical guide to Spring AI: ChatClient, advisors, chat memory, RAG with pgvector,…
Let's talk
Tell me about your AI challenge
Describe what you want to automate or build — I'll reply within one business day with a call agenda and a rough estimate.
Email is the fastest way. A few sentences about the task are enough to start.
By sending a request, you agree to the privacy policy.