Browse categories
Explore
Fiverr Pro
English
$
USD
I will do data science, ml and predictive analytics with python
AI Agent Developer Claude Code LangChain n8n Data Science Expert
MSc Data Science (SZABIST 2024) + IBM Certified Data Scientist.
I build production AI agents and ML systems that ship business outcomes — not demos. 20+ projects: DT International churn (92.85% ROC ...
About this Gig
Your spreadsheet has a decision hiding in it. My job is finding it and proving it holds up.
92.85% ROC AUC churn model (DT International), Prophet R²=0.80 on 160K+ listings, SVM 86% on 4M+ Amazon reviews.
I do data science and ML in advance Python churn, forecasting, classification, segmentation from messy raw data to a model you can trust and act on.
I don't hand back a notebook full of charts and call it done. Every project ends with a plain-English read on what the model found, how confident it is, and what it means for your decision.
What you get:
- Exploratory data analysis + a data-quality report (messy data is normal I plan for it)
- Feature engineering and model selection (Random Forest, XGBoost, Prophet, LSTM, and more)
- Validation and tuning, with accuracy reported honestly including where the model struggles
- SHAP explanations so you can see which factors drive each prediction
- A clean handoff: notebook, written report, or a deployed service your choice
Place your order share your dataset and goal in the requirements form. Want a feasibility check? Message me an anonymized sample.
My Portfolio
FAQ
What ML problems do you handle?
Churn prediction, demand/time-series forecasting, classification, segmentation, regression — areas where I have shipped, verifiable results (92.85% ROC AUC, R²=0.80, SVM 86.01%).
My data is messy and incomplete - is that a problem?
That's the normal starting point. Every project opens with an exploratory pass that surfaces missing values, outliers, and inconsistencies, and you get a written data-quality report before modeling.
How do I know predictions aren't a black box?
SHAP explanations (Standard up) show exactly which features drove each prediction, in plain language -I walk you through it on the handoff call.
Will you tell me if the data isn't strong enough?
Yes - directly and early. I'd rather flag a data limitation in week one than ship a model that looks good in testing and fails in production.
Do you deploy, or just hand over a notebook?
Both — Basic is notebook + report; Standard/Premium can include a deployed service (FastAPI, containerized).
Can you handle large datasets?
Yes — recent work processed 4M+ Amazon reviews on Apache Spark with MLlib. Scale is rarely the limit; data quality usually is.
What if accuracy disappoints after delivery?
That's what revisions are for — I retune, re-validate, and use the SHAP breakdown to show what's driving any gap. Revision scope covers methodology + feature engineering on the data you provide; it can't override fundamental data-quality limits, which I flag *before* modeling begins.
Can my team use the results without a data scientist?
Yes — Standard+ includes a Streamlit demo or a plain-English report; to *query* results conversationally, see my RAG chatbot gig; for a live model API, add the FastAPI endpoint extra.
Response speed, and who owns the work?
I reply within a few hours (UK/EU morning + US-East afternoon overlap, async via Fiverr). You own the notebook, models, and code on delivery; I keep no copies; NDA available.
