How to Get Freelance Machine Learning Projects as a Beginner

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import make_pipeline

texts = ["refund not received", "where is my order", "great product",
         "delivery late again", "love the quality", "payment failed twice"]
labels = ["billing", "delivery", "praise", "delivery", "praise", "billing"]

clf = make_pipeline(TfidfVectorizer(), LogisticRegression(max_iter=1000))
clf.fit(texts, labels)
for msg in ["my refund is pending", "order not delivered yet"]:
    print(msg, "->", clf.predict([msg])[0])
# Real projects need many more labelled examples and a test set.

import pandas as pd

sales = pd.Series([120, 132, 128, 140, 151, 149, 160, 158],
                  index=pd.period_range("2026-03", periods=8, freq="M"))

naive = sales.shift(1)                      # last month's value
moving = sales.rolling(3).mean().shift(1)    # 3-month average
for name, pred in [("naive", naive), ("3-month avg", moving)]:
    mae = (sales - pred).abs().dropna().mean()
    print(f"{name:12} MAE = {mae:.1f}")
print("next month (naive):", sales.iloc[-1])

Sample project: Support message categoriser (public data)

Problem: A small online shop receives many messages daily and
wants them sorted into billing, delivery, product and other.

Approach: cleaned text, TF-IDF features, logistic regression;
compared with a keyword-rule baseline.

Result: model beat the baseline on a held-out test set;
most errors were between 'delivery' and 'other'.

Deliverables: trained model, prediction script, report,
instructions to retrain with new labelled data.

Hi <Name>, I help small businesses sort and analyse their
customer messages using simple machine learning.

For example, I built a categoriser that sorts messages into
billing, delivery and product questions: <link>.

If useful, I can do a small fixed-scope trial on a sample of
your messages and share a short report on what it finds.
Happy to answer any questions. Thank you.

# predict.py - handed over to the client
# Usage: python predict.py messages.csv output.csv
import sys
import joblib
import pandas as pd

def main(inp: str, out: str):
    model = joblib.load("message_classifier.joblib")
    df = pd.read_csv(inp)
    if "message" not in df.columns:
        sys.exit("Input CSV needs a 'message' column")
    df["category"] = model.predict(df["message"].fillna(""))
    df.to_csv(out, index=False)
    print(f"Wrote {len(df)} predictions to {out}")

if __name__ == "__main__":
    main(sys.argv[1], sys.argv[2])

# Message Classifier - Handover

## What it does
Sorts customer messages into billing, delivery, product, other.

## How to use
    pip install -r requirements.txt
    python predict.py messages.csv output.csv

## How well it works
Tested on held-out messages; see report.pdf for metrics and
common errors.

## How to retrain
Add labelled rows to training.csv and run train.py.

## Limitations
Works best on English messages similar to the training data.

Hi <Client>,

Thank you for sharing the project. My understanding:
you want customer messages sorted into a few categories.

Proposed approach:
1. Review a sample of your messages and agree categories
2. Build a simple baseline and a machine learning model
3. Compare both on held-out messages and share a report
4. Hand over a prediction script and instructions

Timeline: <dates>   Price: <amount>, in two milestones
Relevant sample: <link>

Deliverables:
  - Cleaned dataset
  - Trained model + prediction script
  - Evaluation report (metrics vs baseline, common errors)
  - Handover README
Not included: app integration, ongoing maintenance
Data: provided by client by <date>, owned by client
Revisions: one round of changes
Payment: <x>% at start, remainder on delivery

Client: ______  Start: ______  Due: ______

[ ] Scope agreed in writing
[ ] Data sample reviewed
[ ] Baseline built
[ ] Model built and evaluated
[ ] Report drafted
[ ] Client review call
[ ] Final delivery + handover
[ ] Payment received
[ ] Ask for feedback or a testimonial (real, in their own words)