Education management
Manual admin work cut by 80–85%
Laravel · MySQL
I'm Kafi, a senior freelance developer. My data science consulting services turn your data into analysis, forecasts, dashboards and tools, written as clean, reproducible Python code that a colleague, client or reviewer can rerun and check.
Starting from $10/hr
The question written down before any code
Code and environment you can rerun in one command
Plain-language findings, with their limits
Documentation for hand-over or peer review
DATA SCIENCE CONSULTING
Data science consulting services help you answer a question with data: from cleaning and exploring it, through analysis or modelling, to a result you can explain and act on. The deliverable is not a model for its own sake. It is a decision, a forecast, a tool or a report that stands up to questions.
Good consulting starts with the question and the data, not the algorithm. Many problems are solved by careful analysis and a clear chart. When modelling is warranted, I start simple, test honestly and say where results are weak.
For research teams the emphasis is reproducibility. I organise code, environments and data so another person can rerun the analysis and get the same results, and I build workflow tools, but I do not write or run assessed coursework, and the scientific conclusions are yours.
| Reproducible consulting | One-off analysis | |
|---|---|---|
| Rerunning it | One command, fixed environment | Works only on one laptop |
| Questions from reviewers | Every step documented | Hard to explain later |
| Changes in data | Refresh the report automatically | Redo by hand |
| Findings | Limits and uncertainty stated | A single confident number |
| Hand-over | Your team can maintain it | Dependent on the author |
WHAT DOES A CONSULTANT DO?
A data science consultant helps you turn data into a decision: framing the question, checking the data, analysing it and explaining what to do. A general guide to the roles around it:
| Data analyst | Data scientist (consultant) | Data engineer | ML engineer | |
|---|---|---|---|---|
| Focus | Reports on what happened | Finds out why and what to do | Builds the data pipelines | Deploys models |
| Typical output | Dashboards | Analysis, recommendations, prototype models | Clean, current tables | A model running in production |
| Page | — | — | Data engineering | Machine learning |
REPRODUCIBLE BY DEFAULT
For analysis you need to defend, such as research, audits and reports, the code and data layout matters as much as the result. This is the structure I set up, so a colleague or reviewer can reproduce your numbers.
data/raw/Original data, never edited
data/processed/Cleaned data, rebuilt by code
notebooks/Exploration, kept readable
src/Tested functions, so results are repeatable
environment.ymlExact package versions, so it runs next year
tests/Checks that results still match after changes
report/Figures and tables generated from the code
README.mdHow to rerun everything in one command
Example layout · adapted to R, Python, notebooks or pipelines as your field requires
WHAT I DO
For business questions and for research tooling, with the work organised so it can be checked.
What is in your data, what is wrong with it and which questions it can answer.
EDA · Profiling
Churn, conversion, pricing, cohort and revenue analysis tied to decisions.
Analytics
Demand, sales and usage forecasts with honest uncertainty ranges.
Time series
Test design and analysis that avoids the common ways results mislead.
A/B · Stats
Simple, explainable models first, with a baseline, then more complex ones if they earn their keep.
Python · scikit-learn
Automatically refreshed reports and dashboards so analysis does not die in a spreadsheet.
Dashboards
Python or R scripts and small apps built for a specific analysis, documented so you can run them.
Python · R
Survey data, open-ended answers and documents analysed with transparent methods and language models where useful.
NLP · Surveys
Literature screening helpers, data-extraction pipelines, summarisers and Streamlit apps that save a research team time, with a human checking the output.
Research · LLMs
Messy analysis code organised, tested and documented, with environments pinned so results can be rerun.
Reproducible
Merging, de-duplicating and standardising data, with every change logged.
Cleaning
A walk-through so your team can maintain the analysis without me.
Training
Describe what you want to build. I'll reply within a few hours and tell you honestly what it would take.
HOW WE CAN WORK
Cost depends on the question, data quality, the amount of data and the format of the result. Typical ways to start:
A short look at your data and goals, ending with a written recommendation on what is worth doing.
A defined question answered with clean, documented code and a readable summary.
A tested first version so you can judge value before investing more.
Numbers your team reads each week, kept current by a pipeline.
Flexible hours from $10 per hour for questions as they come up.
Decide whether a hosted model, a custom model or neither fits.
HOW IT RUNS
Clear scope up front, review points throughout.
What you need to decide or show, and what would change your mind.
Quality, gaps and what the data can and cannot support.
Methods agreed in writing before running them.
Code written, tested and reviewed, with figures generated by the code.
Sensitivity checks and the strongest objections tried.
Findings, limits and a repository your team can rerun.
HONEST ADVICE
Set up tracking and storage first.
A report is cheaper than a model.
Some problems are workflow problems, not data problems.
A hosted model may already do the job.
Small samples support small claims; I will say so.
Start from what you would do differently with the answer.
TOOLS
No proprietary lock-in.
PROOF
Real products in production, described without client names. Each one started with a conversation like the one you're about to have.
Manual admin work cut by 80–85%
Laravel · MySQL
Personalised kids' books with AI face swapping
Laravel 13 · Fabric.js · Segmind AI
Member, plot and transaction management in 2 weeks
Laravel
COST & TIMELINE
A focused analysis of one dataset is a small project; building forecasts, dashboards and tools around several sources is larger. You get a written scope and estimate after a short call, and hourly work starts at $10 when you hire me directly.
Any cloud, compute or language-model fees are billed to your accounts at cost.
Cleaning is often the largest part.
A sharp question saves days of exploration.
Simple analysis costs less than modelling with validation.
Dashboards, apps and documentation add work.
Work for publication or audit needs extra checking and documentation.
RESEARCHERS AND US TEAMS
For researchers I set up reproducible analysis environments, pipelines and visualisations, and review code. I do not write or run assessed coursework. For US businesses I work in US dollars from $10 per hour with overlap hours and a written scope.
Cost depends on the question, the amount and quality of data, how much cleaning is needed and what you want delivered: a report, a dashboard or a model.
WHO IT'S FOR
Where it pays off fastest.
PRICING & PLATFORMS
Work with me directly for the best rate, or through Upwork or Fiverr if you want the platform's payment protection.
From $10/hr
Upwork
Fiverr
Long-term work starts at $10 per hour. Every project gets a written estimate before it starts, so there are no surprises.
FAQ · 12 QUESTIONS
Still unsure? Book a free call and ask me directly. I'll tell you honestly if I'm the right fit.
They help you answer questions with data: cleaning and exploring it, analysing or modelling it and delivering a result, forecast, dashboard or tool you can explain and act on.
It depends on the data, the question and the deliverables. I give a written estimate after a short call, and hourly work starts at $10 when you hire me directly.
Data engineering builds the pipelines and storage that make data reliable. Data science analyses that data to answer questions or build predictive tools. Many projects need a little of both.
Often not. Careful analysis and a clear chart answer many questions. I try the simple approach first and use ML only if it clearly adds value.
Yes, on tools and code: analysis scripts, data-extraction pipelines, summarisers, Streamlit apps and reproducible project setups. The scientific conclusions stay with the researchers.
No. I help teams and researchers with tools, code quality and reproducibility, not with work that someone is being assessed on.
Yes. Language-model pipelines can screen abstracts and extract structured data from papers. Current research says they need human checking, so I build them with review steps and visible sources.
Yes. I organise the code, pin the environment, add tests and a one-command rerun, and document the data, so another person can reproduce your results.
Mainly Python, with SQL for data. I can work in R where your team already uses it, or tell you honestly if another specialist would suit better.
Yes. Looker Studio, Metabase, Streamlit or a custom dashboard, backed by a pipeline that refreshes on schedule.
I work in your environment where possible, use the minimum access needed and tell you which tools see which data. For regulated data, your compliance rules take priority.
You do. Code, notebooks, data documentation and reports are handed over, and nothing is locked to me.
Tell me what you're building and where you're stuck. You'll get an honest plan, timeline and estimate, even if I'm not the right fit.