AI/ML Internship · Week 4 · Capstone Project
Week 4
Intro
Chapters:
Synkoc AI/ML Internship · Week 4 · Lessons 11-13 of 13
Capstone
Project
Apply everything from 3 weeks in one real-world ML project. Choose 2 from 10 industry-grade projects. Build end-to-end. Present your results.
🌟 Choose 2 Projects
📈 End-to-End ML
📄 Present & Document
🏅 Earn Certificate
🧑‍💻
Synkoc Instructor
AI/ML Professional · Bangalore
⏳ ~60 minutes
🟢 Capstone Level
The Capstone Challenge
Week 4 is your proof of mastery. No more tutorials. You design, build, and present two complete ML solutions independently.
🎯
Choose any 2 from 10 projects.
Each project is a real industry problem. You must build a complete end-to-end ML pipeline: data preparation, model training, evaluation, and documented results. Both projects contribute equally to your final grade.
1
Pick 2 Projects
→
2
Load & Explore Data
→
3
Clean & Engineer
→
4
Train & Tune
→
5
Evaluate & Report
→
6
Present Results
Chapter 1 of 3
01
Projects 1 to 5
Five industry-grade ML projects. Classification, regression, NLP, and computer vision. Pick the ones that excite you most.
Projects 1 — 5
Project 01 · Classification
📩 Spam Email Classifier
Train a model to detect and filter spam emails based on message content using Natural Language Processing.
  • Extract features from raw email text using TF-IDF vectorisation
  • Train Naive Bayes and Logistic Regression classifiers
  • Evaluate with Precision, Recall and F1-Score (spam = positive class)
  • Explain why high Recall matters more than Precision for spam
NLPTF-IDFNaive BayesClassification
Project 02 · Regression
🏠 House Price Predictor
Predict residential property prices using location, size, amenities, and neighbourhood features.
  • Perform full EDA: distributions, correlations, outlier detection
  • Engineer features: price per sqft, age of property, distance to centre
  • Compare Linear Regression vs Random Forest Regressor
  • Report MSE, RMSE, R2 and interpret which features drive price
RegressionEDAFeature Eng.Random Forest
Project 03 · Classification
💊 Credit Card Fraud Detector
Identify fraudulent transactions in a heavily imbalanced dataset where only 0.17% of transactions are fraud.
  • Handle extreme class imbalance using SMOTE oversampling
  • Train Random Forest and XGBoost on the resampled data
  • Demonstrate why accuracy is misleading: use ROC-AUC and F1
  • Set decision threshold to minimise false negatives (missed fraud)
Imbalanced DataSMOTEROC-AUCXGBoost
Project 04 · Clustering
👥 Customer Segmentation
Group e-commerce customers by purchasing behaviour to enable targeted marketing campaigns.
  • Compute RFM features: Recency, Frequency, Monetary value per customer
  • Apply Elbow Method to choose optimal K, then run KMeans clustering
  • Use PCA to visualise clusters in 2D scatter plot
  • Name and describe each segment: Champions, At-Risk, New Customers
RFM AnalysisKMeansPCAMarketing
Project 05 · NLP
🥳 Sentiment Analyser
Classify product reviews as Positive, Neutral, or Negative to help businesses monitor customer satisfaction.
  • Clean and preprocess text: lowercase, remove stopwords, lemmatise
  • Vectorise with TF-IDF and train multi-class Logistic Regression
  • Build a prediction function that takes any review string as input
  • Report per-class F1 and identify which sentiment class is hardest to predict
NLPText CleaningTF-IDFMulti-class
Chapter 2 of 3
02
Projects 6 to 10
Five more projects covering medical AI, time-series, recommendation systems, image classification, and energy prediction.
Projects 6 — 10
Project 06 · Medical AI
💊 Diabetes Risk Predictor
Predict whether a patient is at risk of diabetes using clinical measurements such as glucose, BMI, and age.
  • Handle missing values coded as zeros (physiologically impossible readings)
  • Perform feature importance analysis to identify top clinical predictors
  • Train and compare Logistic Regression, Decision Tree, and Random Forest
  • Justify why Recall matters more than Precision in medical diagnosis
Healthcare AIClinical DataFeature Importance
Project 07 · Time Series
📈 Sales Forecasting
Predict future weekly sales for a retail chain using historical transaction data and seasonal patterns.
  • Engineer time features: day of week, month, is_holiday, rolling averages
  • Train Random Forest Regressor on the engineered time features
  • Evaluate with MAE and MAPE (Mean Absolute Percentage Error)
  • Visualise predicted vs actual sales on a time-series line chart
Time SeriesFeature Eng.RetailForecasting
Project 08 · Recommender
🎬 Movie Recommendation Engine
Build a content-based recommendation system that suggests similar movies based on genre, cast, and plot keywords.
  • Combine genre, cast, director and keyword text into one feature string
  • Compute TF-IDF vectors and cosine similarity between all movie pairs
  • Build a function: input any movie title, return top 10 similar movies
  • Extend with collaborative filtering using user rating matrix
RecommenderCosine SimilarityTF-IDFContent-Based
Project 09 · Computer Vision
📷 Handwritten Digit Classifier
Classify handwritten digits 0-9 from 28x28 pixel images using the MNIST dataset — the Hello World of deep learning.
  • Flatten images from (28,28) to (784,) feature vectors, normalise 0-1
  • Train Random Forest and compare to a simple Neural Network with sklearn
  • Display a confusion matrix heatmap to see which digits are confused
  • Visualise misclassified samples and explain why the model failed
Computer VisionMNISTImage DataNeural Net
Project 10 · Regression
⚡ Energy Consumption Predictor
Predict hourly electricity demand for a city using weather, time, and economic activity data.
  • Engineer cyclical time features: encode hour and month as sin/cos pairs
  • Handle weather outliers and missing sensor readings
  • Compare Gradient Boosting vs Linear Regression for energy forecasting
  • Report feature importances and explain which factors drive consumption peaks
Energy AICyclical FeaturesGradient BoostingRegression
Chapter 3 of 3
03
How to Build Your Capstone
The 6-step professional workflow every data scientist follows. Your capstone must demonstrate all 6 steps for both projects.
The 6-Step Build Process
Every project follows this structure. Each step builds on the previous. Skip none.
Step 01
📊 Data Loading & EDA
Load dataset. Print shape, dtypes, head(). Plot distributions. Check missing values. Compute correlation matrix.
  • df.info(), df.describe(), df.isnull().sum()
  • At least 3 visualisation charts
  • State 3 key observations from EDA
Step 02
🧹 Data Cleaning & Features
Handle missing values, encode categoricals, create new features, normalise numeric columns.
  • fillna with median/mean/mode
  • LabelEncoder or one-hot for categoricals
  • At least 1 engineered feature
Step 03
⚙ Model Training
Train at least 2 different algorithms. Use train_test_split with random_state=42. Apply cross-validation.
  • Minimum 2 algorithms compared
  • cross_val_score with cv=5
  • Print mean +/- std for each model
Step 04
📈 Evaluation
Use the right metrics for your problem type. Classification: F1, ROC-AUC. Regression: MSE, R2.
  • classification_report or MSE+R2
  • Confusion matrix for classification
  • ROC-AUC with predict_proba
Step 05
🔎 Insights & Analysis
Print feature importances. Explain which variables matter most. Discuss model limitations.
  • feature_importances_ bar chart
  • At least 3 business insights
  • State what would improve the model
Step 06
📄 Report & Present
Write a structured report. Present findings in clear language a non-technical stakeholder can understand.
  • Executive summary (3 sentences)
  • Results table: model vs metric
  • Recommendation for production deployment
Grading Rubric
Each project is graded out of 50 points. Two projects = 100 points total. Pass threshold: 70 points.
15 pts
📊 EDA & Cleaning
Thorough EDA with charts, correct handling of missing values, feature engineering, data is ready for training
20 pts
⚙ Model & Evaluation
Minimum 2 models compared, cross-validation used, correct metric for problem type, results clearly reported
15 pts
📄 Report & Insights
Executive summary, feature importance analysis, at least 3 business insights, deployment recommendation
+5 bonus
🌟 Innovation
Bonus points for creative feature engineering, hyperparameter tuning, or going beyond the minimum requirements
70/100
🏅 Pass Threshold
Score 70 or above across both projects to earn the Synkoc AI/ML Internship Certificate of Completion
Jupyter
📄 Submission Format
Submit as .ipynb Jupyter Notebook with all cells executed, or Python .py file with full comments and output
Tips for Success
From Synkoc instructors who have reviewed hundreds of capstone projects. Do these and you will stand out.
🎯
Choose What Interests You
Pick projects from domains you find genuinely interesting. Enthusiasm shows in the quality of analysis. Your best work comes from real curiosity.
📊
Spend 40% on EDA
Most students rush to modelling. The best projects spend the most time understanding the data. Every insight you find in EDA makes the model better.
⚙
Always Compare 2+ Models
Never submit a project with only one model. Always compare at least two using cross-validation. This shows you know how to make evidence-based decisions.
🔎
Explain Your Choices
Why did you choose this metric? Why this model? Explaining your reasoning clearly is what separates junior from senior data scientists.
👥
Write for Non-Technical Readers
Your executive summary should be understandable by a business manager who knows nothing about ML. This is the most valuable skill in industry.
🌟
Go Beyond the Minimum
The minimum gets you a pass. Going beyond, with hyperparameter tuning, creative features, or extra visualisations, gets you a distinction and impresses future employers.
Everything You Have Mastered
In 4 weeks you covered the complete ML engineering stack. Your capstone is proof of mastery.
Week 1
🔨 Foundation
Python variables, loops, functions, data structures. Statistics: mean, std, probability, correlation, EDA.
Week 2
📈 Data Tools
NumPy arrays and vectorised ops. Pandas DataFrames, cleaning, groupby. Matplotlib and Seaborn visualisation. EDA pipeline.
Week 3
🤖 Machine Learning
Supervised: Linear Reg, Logistic Reg, Decision Tree, Random Forest. Unsupervised: KMeans, PCA. Evaluation: F1, CV, ROC-AUC.
Week 4
🌟 Capstone
End-to-end ML on real data. Choose 2 from 10 real-world projects. Build, evaluate, document, and present complete solutions.
🏅
Ready to Build?
You have the knowledge. You have the tools. Now prove it. Choose your 2 projects from the Capstone Lab, build complete pipelines, and earn your Synkoc AI/ML Internship Certificate.
✓ 3 Weeks of Training Done
✎ Choose 2 Projects — Now
🏅 Certificate on Completion
Good luck from the entire Synkoc team in Bangalore. We believe in you.
Synkoc IT Services · support@synkoc.com · +91-9019532023
Press ► Play to start the lesson with voice narration
0:00 / ~60:00
🔊
1 / 12