Łukasz Marchlewicz

Aspiring Data Scientist / AI Engineer with 15+ years of business experience

🧭 About

I am a data-driven professional transitioning from entrepreneurship into Data Science, Machine Learning and AI.

For over 15 years I ran my own stone processing and manufacturing business in Poland, working with B2B clients and international partners. I was responsible for:

developing and maintaining long-term B2B relationships,
pricing, cost analysis and profitability of projects,
coordinating production, logistics and quality control,
negotiating and delivering contracts end-to-end.

Alongside this, I am also building a second business in visual media –
aerial cinematography and video production with FPV drones – under the brand:

natural stone : lumargo.pl
filmmaking & drone studio: studiofilmowania.pl

Over time I realised that many of my business decisions were already data-driven:
I regularly analysed sales trends, seasonality, margins and operational costs in Excel.
This naturally led me towards Data Science and AI, where I can combine:

practical business understanding,
analytical thinking,
and modern tools such as Machine Learning and Large Language Models (LLMs).

Today my goal is to grow into roles such as:

Junior Data Scientist,
Junior / Associate Data Analyst,
AI / LLM Engineer (entry-level),

where I can bring both hands-on ML skills and real business experience from running a company.

🧮 Technical Profile (in progress)

I am actively developing my technical skill set and focusing on end-to-end projects rather than only theory.

Languages & Tools

Python – data analysis, ML prototyping
SQL – querying of relational data
Advanced Excel – financial / operational analysis
GitHub – version control and project documentation
n8n – low-code orchestration and connecting AI agents with external tools

Machine Learning (junior level, growing)

Supervised learning: classification & regression
Model workflow: data cleaning, feature engineering, training, validation, metrics
Libraries: pandas, numpy, scikit-learn, xgboost, catboost
Experiment tracking: basic use of MLflow

NLP & LLMs

Text cleaning, tokenization, TF-IDF and classical ML models for text
Practical use of LLM APIs (OpenAI, Gemini) to build:
- simple chatbots,
- AI agents with tools (tool calling),
- prototypes of RAG (Retrieval-Augmented Generation) over custom knowledge bases.

I am fully aware that my technical journey is still at an early stage, but I am compensating with:

consistent, structured learning,
building real projects,
and strong motivation to grow in Data & AI.

📁 Portfolio

Below are projects that I am currently developing and documenting.
They are designed as realistic, business-oriented use cases where I can apply ML and LLMs end-to-end.

1. Local Expense Analysis – Transaction Categorization

Goal

Build a local, privacy-friendly tool that analyses personal or small-business bank statements and automatically classifies expenses into meaningful categories.

What I am working on

Parsing and cleaning transaction data (CSV / XLS exports from banks) with Python & pandas
Designing rule-based mappings for merchants and transaction descriptions (e.g. merchant → category)
Experimenting with ML models to classify transactions based on text description, amount and merchant
Aggregating results into simple reports: monthly spending by category, trends, top merchants

Tech stack

Python, pandas, scikit-learn
Local CSV files, Jupyter notebooks
First experiments with MLflow for tracking model versions

Why this project matters

It reflects a real-world use case that I personally face as an entrepreneur
It combines data cleaning, feature engineering, ML modelling and reporting
It is a solid practical exercise in building an end-to-end pipeline on local data

2. AI Agents for Clients – Product Search & Knowledge Assistant

Goal

Design an AI agent that can support customers of a small/medium business by:

searching over a structured product catalog (e.g. stones, headstones, countertops),
answering detailed questions from a knowledge base (materials, dimensions, installation, maintenance),
integrating with a calendar to propose and schedule consultations.

What I am working on

Defining the orchestrator agent prompt and decision logic
Implementing tools (API endpoints) for:
- search – product search with filters (category, material, color, etc.),
- answer – question-answering over a knowledge base (documentation, FAQs),
- calendar tools – reading and creating events for meetings with clients.
Integrating everything inside n8n:
- chat/agent nodes using OpenAI / Gemini,
- HTTP Request nodes (Search & Answer),
- structured workflows combining multiple agents (calendar + knowledge).
Experimenting with RAG: storing company knowledge in a vector database and using LLMs to generate helpful, grounded answers.

Tech stack

n8n (self-hosted orchestration)
OpenAI / Gemini APIs for LLM reasoning
HTTP endpoints for search & answer tools
Vector database for knowledge retrieval (e.g. Pinecone / Chroma – under evaluation)

Why this project matters

It matches my real business context (stone products and services for clients)
It shows how LLMs and agents can bring value to traditional industries
It is a good foundation to grow into AI Engineer / LLM Engineer roles

🎓 Education & Training

Formal Education

Postgraduate Studies – Data Science in Business
Nicolaus Copernicus University, Toruń
MSc – Management Information Systems
Nicolaus Copernicus University, Toruń

Selected Courses

DataWorkshop – Practical Machine Learning
DataWorkshop – Natural Language Processing (NLP)
DataWorkshop – Practical LLMs
- building LLM-powered apps,
- RAG and agents,
- ML pipelines in practice.

I treat these as a starting point, not as proof of being “finished”.
I am continuously revisiting the material and re-implementing concepts in my own projects.

🚀 What I am looking for

I am currently looking for opportunities where I can:

contribute to data-driven decision making,
grow under more experienced Data Scientists / ML Engineers,
and gradually take ownership of end-to-end analytical or AI projects.

I am especially interested in:

Junior Data Analyst / Junior Data Scientist roles,
early-stage AI/ML teams that work with LLMs and agents,
companies that value both technical skills and business thinking.

📫 Contact

The best way to reach me is via:

GitHub – github.com/lukasmarch
LinkedIn – linkedin.com/in/łukasz-marchlewicz-0a832a169

If you are hiring for Data / AI roles (junior or transition profiles) or looking for collaboration on ML / LLM projects,
I would be happy to connect.