Skip to content
EP.
Open to roles in AI automation

Evgenii
Ponomarev

Business Controller at Timac Agro Slovakia, Bratislava

I automate finance processes and build tools with AI agents.

I run controlling for two entities . I automate recurring Excel work with Python, VBA and various AI tools, and in my free time I build my own apps. I believe the future belongs to AI automation and AI agents, and that is where I want to work.

Evgenii PonomarevAI automationAI agentsAI in finance
processes moved from manual Excel to Python and VBA in 5 months
of manual work saved per month
AI apps I built myself: AgroŠtít and CherryPicker
  • Python
  • Claude Code
  • MCP
  • TypeScript
  • React Native
  • FastAPI
  • pandas
  • VBA
  • SAP Business One
  • Power BI
  • Power Query
  • whisper.cpp

About

Since May 2026 I have been Business Controller at Timac Agro Slovakia, part of Groupe Roullier . I handle month-end closes and group reporting for the Czech and the Slovak entities, and help the CFO get a clearer view of how the business runs.

I started by automating my own work, then built tools for colleagues in logistics, HR and office administration. Every tool is easy to use, properly documented and verifiable: its output has to reconcile to a control total with zero difference.

Before that I worked in finance at Deutsche Telekom Cloud Services and Arval Slovakia, built pricing models for the staffing company ATALIS, and co-founded Coffee Corner, a network of self-service coffee stations, where I owned the P&L and cash flow.

Based in
Bratislava, Slovakia
Languages
Slovak C2, English C1, Russian (native)
Studies
MSc Financial Management, Comenius University (2027)
Tools
Python, TypeScript, VBA, SQL, Claude Code, MCP, SAP Business One, Power BI
LinkedIn profile

Experience

  1. 2026 – now

    Business Controller (SK and CZ)

    Timac Agro Slovakia, Groupe Roullier

  2. 2025

    Pricing models for staffing deals

    ATALIS, contractor

  3. 2024 – 2026

    Financial data analyst

    Deutsche Telekom Cloud Services

  4. 2023 – 2025

    Co-founder, finance and P&L

    Coffee Corner, self-service coffee stations

  5. 2023

    Accountant assistant, receivables

    Arval Slovakia

How I work with AI

  • Human in the loop

    Where money, customers or personal data are involved, a person always has the final say. AI drafts e-mails and payments for approval, it does not send them.

  • People first, then code

    I start by talking to the people who do the work and fit the tool to their process, not the other way round.

  • Verifiable results

    Control totals, golden tests and evals instead of a feeling that it looks right.

  • Data protection

    I send only the data that is needed, mask personal data and transcribe speech locally where possible.

  • Test before go-live

    A new tool runs in test mode or in parallel with the old process first.

Projects

Outside work I build my own apps in React Native, write TypeScript and Python, and use AI agents throughout development. Here are two of them, plus the tools that run in my day job every week.

AgroŠtít on a phone at a kitchen table next to an invoice

AgroŠtít

A mobile app that uses AI to check a suspicious invoice, text message, e-mail or phone call. It is made for farmers and older people in Czechia and Slovakia.

iOS and Android in closed testing. Entered in the AI výzva pro Česko 2026 competition.

eval cases
automated tests
input types: invoice, text, e-mail, call
Home screen: check a message
Check result with warning signs
  • Expo / React Native in TypeScript with a FastAPI backend. Claude returns a risk level, the warning signs it found, quoted, and what to do next.
  • Personal data is masked before anything is sent. Outputs follow a JSON Schema, companies are checked against the Czech VAT registry and ARES, and inputs are screened for prompt injection.
  • Quality is guarded by evals with an LLM judge and automated tests. Speech-to-text runs on a self-hosted whisper.cpp, the backend on an EU VPS.
CherryPicker on a phone next to promo leaflets and vegetables

CherryPicker

A week of dinners planned from the promo leaflets of Lidl, Kaufland, Tesco and Billa.

iOS on the App Store, Android in closed testing, web app.

promotions from leaflets
grocery chains
weekly plans precomputed
CherryPicker on the App Store
Weekly meal plan from promotions
Shopping list by store
  • A Python pipeline (pdfplumber, pydantic) pulls promotions out of PDF leaflets and matches products to recipes using rules and cached LLM decisions.
  • An Expo / React Native app for iOS, Android and the web. Basket maths lives in both Python and TypeScript, and a golden test keeps them equal to the cent.
  • An MCP server with six read-only tools so AI agents can use the planner. A 15-case tool-selection eval catches problems and bugs in how agents use it.

Finance process automation

Python and VBA tools that my colleagues and I use at Timac Agro. Figures come from my work there since May 2026.

  • “E-reminders”: a Python pipeline that checks, generates and sends payment reminders to customers

    Before
    4–5 h a week
    Now
    ~15 min

  • VBA macro that calculates sales team commissions in two countries (~40 reps)

    Before
    1–1.5 days
    Now
    3–4 h

  • Pipeline that checks business trips for HR, analysing 4,700+ GPS trips

    Before
    1 day a month
    Now
    1 click

  • Credit controlling

    Before
    2 h per cycle
    Now
    20 min

  • Work procedures (SOPs) from recordings: local speech-to-text and an LLM draft

    Before
    5–8 h each
    Now
    0.5–1 h

  • Automated comparison and check of sales visits: CRM against GPS

    Before
    53 unexplained
    Now
    21

Python (pandas, python-docx), VBA, Power Query, SAP Business One, an LLM review agent

Calculators

Four calculators for self-employed people in Slovakia. Everything is calculated in your browser.

Let’s talk

If you would like to talk about a role or one of the projects, write to me.