Upwork home
Sign up

Development & IT Consultation with Taras Z.

4.9 · 13 reviews

Development & IT Consultation with Taras Z.

4.9 · 13 reviews

AI Strategy & Architecture Consult (Agents + RAG)

Building a RAG assistant or tool-using agents? I’ll help you turn ideas into a ship-ready plan: what to build, how to build it securely, and how to measure success.

We can cover:
• RAG blueprint: data sources, chunking, retrieval/reranking, grounded answers + citations, anti-hallucination guardrails
• Agent design: sequential/parallel/hierarchical flows, tool calling, routing/fallbacks, safe execution
• Cost & latency: long-context optimization (summarization/compression/selective retrieval), token spend control
• Security/privacy: local/on-prem constraints, data handling, risk controls
• Evaluation: benchmarks, regression tests, error taxonomy

You’ll leave with: an architecture recommendation, a prioritized 30–90 day roadmap, and a clear next-steps checklist.

Bring your target workflow, constraints (budget/latency/security), and (if possible) sample docs.

PhD in NLP/ML, 15+ years practical experience.
Get personalized advice on:
AI & Machine Learning AI Integration Chatbot Development Prompt Engineering ai agent development

You’re covered with payment protection

Taras Z.Status: Offline

About Taras

Taras Z.Status: Offline
AI / NLP Engineer (PhD): Agentic AI, RAG, Long-Context Systems
100% Job Success
4.9  (13 reviews)
Lahti, Finland - 12:10 am local time
🎓 PhD in NLP and ML
💬 Agents, Chatbots, GPT, and LLMs
📊 Generative AI, Agentic Systems, Sentiment Analysis, Text Classification, and Data Clustering
📸 Non-textual data

I build production-ready RAG + agentic workflows: grounded answers with citations, tool-using agents, routing/fallbacks, and evaluation/regression testing. Strong focus on long-context optimization (summarization, compression, selective retrieval) to cut token cost and improve reliability.

If you have internal documents, knowledge bases, tickets, contracts, or product docs—and you want an assistant that can retrieve the right evidence, reason over it, cite it, and take actions via tools—I can design and implement the full solution end-to-end.

What I deliver

- RAG done right (not just embeddings): chunking strategy, metadata, retrieval tuning, reranking, query rewriting, citation-grounded answers, and “no hallucination” guardrails.
- Agentic workflows: hierarchical (manager→specialists), parallel specialists, and sequential pipelines using LangChain/LangGraph (and CrewAI when it fits).
- Tool-using agents: API/tool calling, safe execution, structured outputs, validation, and robust fallback behavior.
- Long-context optimization: hierarchical summarization, context compression, selective reading, heuristic truncation—reducing token cost while improving stability on large inputs.
- Evaluation & reliability: task benchmarks, regression tests, error taxonomy, confidence routing, and human-in-the-loop review flows.

Typical client outcomes

- Higher answer quality with grounded citations
- Lower costs via routing + compression + selective retrieval
- A system you can trust in production: measured accuracy, predictable behavior, and clear failure modes

What it’s like to work with me

- I start with your success criteria (accuracy, latency, cost, security) and turn it into a concrete plan.
- I communicate clearly, ship incrementally, and keep scope under control.
- I can work within strict constraints (privacy/security, “no external API calls,” on-prem/local models).

🎓 PhD in NLP/ML (UK).
If you share your data shape + target use case, I’ll propose the fastest path to a reliable RAG/agent system—with measurable milestones.

What to expect

Schedule the consultation
Choose from the freelancer’s available days and times.
Get advice for your custom needs
Share details about your project and what you want to talk about. The freelancer will review and reach out if they have questions.
Join the Zoom meeting
1-on-1 meeting with the freelancer to discuss your needs and project.
Approve the work
The freelancer will finish up the documents you asked for and send them to you for approval:
    Before the consultation

    Here’s what Taras will need to know before you meet

    1. Tell me about yourself and what you want to talk about. Share as much of your data as it makes sense so that I can better understand your needs.
    Rating is 4.9 out of 5.
    (13)
    5 stars
    92% Complete
    (12)
    4 stars
    8% Complete
    (1)
    3 stars
    1% Complete
    (0)
    2 stars
    1% Complete
    (0)
    1 star
    1% Complete
    (0)
    RL
    Ronny L.
    5.0
    Mar 19, 2026
    Natural Language Processing Expert Taras has been instrumental in delivering several of Kontrad’s most critical technical components. He led the development of our document extraction pipeline, successfully preserving complex document structures, and played a key role in implementing clause labelling and our core AI-driven feature for participant input evaluation.

    Beyond his technical capabilities, Taras is highly engaged and proactive. He's collaborated with both backend and frontend teams, and consistently brings thoughtful, alternative approaches when tackling complex problems. His ability to combine strong engineering fundamentals with practical problem-solving has been a major asset to the project.

    He is also a pleasure to work with on a personal level - reliable, knowledgeable, and genuinely committed to the success of the product.

    I would strongly recommend Taras for any project involving data engineering, NLP, or applied AI systems.
    LS
    Lutz S.
    4.0
    Oct 1, 2025
    Expert help in langchain / LLM chatbot creation
    MR
    Michael R.
    5.0
    Apr 16, 2025
    Expert NLP/AI Developer Taras is the best NLP developer I've worked with on UpWork. He solved incredibly complex challenges and a tough environment and was a winning factor on the team. Easiest recommendation to anyone who wants to work with him.
    IM
    Irene M.
    5.0
    Nov 1, 2023
    Back-End Development | Project Wrap Up
    AL
    Alex L.
    5.0
    Jul 6, 2023
    Build components over GPT to improve the legal workflow Taras's communication and turnaround time is spot-on and he is a professional teammate. He listened to feedback well iterates quickly.