Browse categories
Explore
Fiverr Pro
English
$
USD
I will build machine learning and forecasting models in python
Data Scientist, RAG and LLM Engineer
Data Scientist and AI Engineer, based in Paris, with two master's degrees in AI (University of Luxembourg and DataScienceTech Institute) and Neo4j certification.
I cover the full pipeline: web scrapi...
About this Gig
A model that cannot beat a simple baseline is worse than no model, because it looks like progress. I check that first, and tell you if the answer is no.
I built demand forecasting and dynamic pricing for a vehicle retailer across several regions, and predictive models on core banking data at a Luxembourg private bank.
WHAT YOU GET
Your data cleaned, with the problems in it named rather than quietly patched
The right model for the problem: forecasting, classification, regression, or clustering
Accuracy measured on a holdout period against a naive baseline, so you see real error, not a claim
Clear explanation of what drives the predictions
Excel, Power BI or API output your team can use without touching Python
Commented code and a walkthrough
STACK
Python, scikit-learn, XGBoost, Prophet, ARIMA, PyTorch, Pandas, SQL, Power BI.
WHY ME
Two master's degrees in AI. Forecasting, pricing and banking analytics delivered for real clients, not just notebooks.
BEFORE YOU ORDER
Send me a sample of your data and tell me what decision it should inform. Some datasets are not ready to model, and I would rather say so than take the order.
My Portfolio
FAQ
How much data do I need?
For forecasting, two full years of history lets the model separate seasonality from trend; twelve to eighteen months can still work with wider confidence intervals. For classification, a few thousand labelled rows is a reasonable floor. Send a sample and I will tell you honestly what is achievable.
How accurate will the model be?
That depends on your data, and anyone quoting a number before seeing it is guessing. What I commit to is measuring it properly: I hold out a recent period the model never sees, predict it, and report the real error against a naive baseline. If the model cannot beat that baseline, I will say so.
What format should I send my data in?
Excel, CSV, or a direct database connection all work. It does not need to be clean; messy exports are normal and cleaning is part of the job. Tell me what each column means, or send whatever documentation exists, and I will work from that.
Do I need to know Python to use it?
No. Deliverables land in Excel, Power BI or an API, so your team works in tools they already use. The commented code comes with it in case you later want to bring it in-house, but nothing about using the output requires touching it.
Can the model keep running on new data?
Yes. Premium includes a refresh pipeline that pulls new data and regenerates predictions on a schedule, so results stay current without anyone rerunning anything. I can deliver it as a scheduled script, a Docker container, or an AWS job depending on your setup.
