Skip to content

01 · TRIGGER — AI Engineer · Agentic AI Developer

Maciej
Małachowski

I wire language models into systems that do real work.

AI developer building LLM systems, AI agents, and end-to-end AI solutions — agents that reason over real data, call real tools, and run in production, not in demos.

View work Download CV
Open for AI work

02 · PROC — Experience

Where I’ve shipped

  1. AI Developer · autoMEE

    Mar 2025 — Present

    Designing and shipping production AI: an accounting-document workflow platform and tailor-made agentic automations integrating LLMs with ERP systems and external APIs.

    • Developing the core of Flowmee, an accounting document workflow platform — document-processing logic, process automation, and handling of data extracted from invoices, CMR and accounting documents.
    • Designing and delivering integrations with client ERP systems — technical client communication, implementation analysis, data mapping, integration logic, and workflow construction primarily in n8n.
    • Building tailor-made solutions fit to clients’ operational, document and accounting processes — automating data flow between systems, validating documents, and streamlining business teams’ work.
    • Designing AI logic for document processing and data extraction, validating accounting values, and reconciling data from source documents against entries in ERP systems.
    • Using AI-assisted development tools like Claude Code and Codex to speed up prototyping, refactoring, testing, and shipping production features.
    • Implementing advanced automations across document handling, systems integration, data processing, client communication, and operational process optimization.
    • Managing context when working with LLMs — structuring prompts, breaking tasks into stages, and designing stable flows for complex business processes.
    • Python
    • TypeScript
    • OpenAI API
    • Claude
    • Claude Code
    • Codex
    • n8n
    • REST API
    • Playwright
    • Docker

  2. Junior Data Scientist · MLJAR

    Jun 2024 — Oct 2024

    Contributed to mljar-supervised, the open-source AutoML library — improving reliability of core modules and authoring developer-facing documentation.

    • Co-developing the MLJAR-supervised library (AutoML) — optimizing training pipelines and stabilizing system components.
    • Diagnosing and fixing bugs in the library’s core modules, improving the reliability of the open-source software.
    • Authoring and maintaining technical documentation for users and developers.
    • Building AutoML use-case examples on real-world datasets.
    • Writing technical articles on AutoML and modeling best practices.
    • Python
    • Scikit-learn
    • Pandas
    • NumPy
    • mljar-supervised
    • Git

03 · PROC — Client work

Running in production

Systems built for autoMEE clients, in daily use. Confidential engagements — no public repositories, so the details below are the proof.

PROC · autoMEE

Sea-Freight Rate Search & Comparison Platform

Scraping agents and an AI layer that automate finding, comparing, and sharing sea-freight rates across carriers.

  • Designed a multi-carrier platform that scrapes, normalizes, and stores freight rates in a database.
  • Built scraping agents that pull rates from carrier sources and persist them for downstream querying.
  • Created an AI layer for searching, analyzing, and presenting rates across a large volume of offers.
  • Replaced manual, multi-source rate checking with an automated pipeline.
  • Python
  • Playwright
  • Claude
  • OpenAI
  • n8n
  • Supabase
PROC · autoMEE

Invoice & CMR → Excel Extraction System

Extracts data from carrier invoices and CMR documents into structured Excel templates.

  • Built a system that extracts data from carrier invoices and CMR documents and transforms it into structured Excel templates.
  • Designed extraction, mapping, and normalization logic to feed downstream operational and accounting processing.
  • Extended it with a foreign-invoice module handling varied document formats and data structures.
  • Cut manual data re-entry and accelerated file preparation for the client.
  • Python
  • OpenAI
  • n8n
PROC · autoMEE

Accounting Data Validator for ERP

AI validation that cross-checks hundreds of accounting rows against source documents before ERP import.

  • Designed and shipped AI logic that validates client-submitted accounting data against its source attachments.
  • Compares hundreds of rows of accounting data against values on invoices and source documents.
  • Automatically flags discrepancies between ERP entries and source documents, reducing human error.
  • Surfaces potential mismatches to operational teams before the final import into the system.
  • Python
  • OpenAI
  • n8n

04 · PROC — Personal builds

Open source

OSS · 2026

AI Budget Assistant

Bank-statement CSVs → categorized, queryable spending — powered by Claude.

A full-stack personal finance app. A headerless, position-based CSV importer auto-detects each bank’s layout; categorization is rules-first (learned merchant map) with Claude Haiku batched only for unknowns. A plain-language Q&A panel answers questions via tool-calling over read-only query functions executed through a SELECT-only database role — the LLM never writes SQL. Built with strict TypeScript, pure I/O-free domain logic, and a Vitest suite.

  • Next.js 15
  • React 19
  • TypeScript
  • Supabase
  • Claude (Anthropic SDK)
  • Tailwind v4
  • Zod
  • Vitest
Source
OSS · 2025

SmartMailer

A multi-LLM agent that automates email, data retrieval, and record-keeping.

A Python agent that sends personalized emails, queries company data, and manages records. GPT-4 Turbo drives decision-making while a local Ollama Llama model handles data analysis; LlamaIndex provides natural-language access over the data, with Pandas, smtplib and PyYAML wiring the automation together for minimal-touch workflows.

  • Python
  • OpenAI API
  • Ollama / Llama
  • LlamaIndex
  • Pandas
Source
OSS · 2024

Docs AI Chatbot

Chat with your PDF and CSV files using a fully local LLM stack.

A retrieval-augmented chatbot for contextual interaction with documents. GPT4All (via LangChain) reads PDF and CSV content and a Mistral 7B model answers questions over it — all served through a clean Streamlit interface and packaged with Docker for reproducible local deployment.

  • Python
  • LangChain
  • GPT4All
  • Mistral 7B
  • Streamlit
  • Docker
Source

05 · STORE — Toolkit

What I work with

Languages
  • Python
  • JavaScript
  • TypeScript
AI & LLM Systems
  • Agentic AI
  • RAG
  • Prompt Engineering
  • LangChain
  • LlamaIndex
  • OpenAI API
  • Claude
  • Gemini
  • Hugging Face
  • Claude Code
  • Codex
  • Hermes Agent
ML & Data
  • Scikit-learn
  • mljar-supervised
  • Pandas
  • NumPy
Automation & Integrations
  • REST API
  • Webhooks
  • n8n
  • Zapier
  • GoHighLevel
  • Voiceflow
Backend & Apps
  • Next.js
  • Flask
  • Streamlit
  • Supabase
  • Docker
  • Playwright
Tools
  • Git
  • GitHub
  • Linux

06 · CTX — About

Engineer, not enthusiast

I design and ship AI end to end: LLM systems, RAG, and multi-step AI agents that reason, use tools, and plug into the platforms businesses already run on — from external APIs to ERP systems. My day-to-day is turning fuzzy problems into reliable, testable software.

Before AI engineering I contributed to open-source AutoML tooling at MLJAR, and I bring a data-science foundation — math modeling, ML, and clean data handling — to every system I build.

Based in
Gdańsk, Poland
Languages
Polish (native) · English (B2)

Education

  • BSc, Mathematical Modeling & Data Analysis

    University of Gdańsk

    2023 — 2026 (expected)

  • Programming Technician

    ZSE im. prof. J. Groszkowskiego

    2019 — 2023

07 · OUT — Contact

Have a problem for an agent?

Open to AI engineering roles and to building agentic, LLM-powered products. Email reaches me fastest.

Email me GitHub LinkedIn

maciejmalachowski1@gmail.com