A complete portfolio of 4 hands-on AI projects built during the DecodeLabs AI Engineering Internship (Batch 2026) — progressing from rule-based logic to supervised learning, recommendation systems, and computer vision.
| # | Project | Core Concept | Tech Stack |
|---|---|---|---|
| 1 | Rule-Based AI Chatbot | Control flow & decision logic | Python, Dictionaries |
| 2 | Data Classification | Supervised learning (KNN) | Scikit-Learn, Iris Dataset |
| 3 | AI Recommendation Logic | Content-based filtering | TF-IDF, Cosine Similarity |
| 4 | Image & Text Recognition | Computer vision & OCR | OpenCV, Tesseract, MobileNet- |
| SSD |
Goal: Build a chatbot that responds to predefined inputs using pure if-else/dictionary logic — no ML involved.
- Continuous
while Trueinput loop with a clean exit strategy - Input sanitization (
.lower().strip()) - 20+ intent knowledge base with keyword-fallback matching
- Demonstrates: Control flow, decision-making logic, basic AI concepts
📁 Project1_RuleBased_Chatbot/ → chatbot.py
Goal: Train a supervised classification model on a real dataset and evaluate it properly.
- Dataset: Iris (150 samples, 3 classes, 4 features)
- Pipeline: Feature scaling (
StandardScaler) → Train/Test split (80/20) →KNeighborsClassifier(K=5) - Evaluation: Confusion matrix, F1 score, full classification report
- Result: 100% accuracy, F1 = 1.0000 on the test set
- Demonstrates: Data handling, supervised learning basics, model training
📁 Project2_Data_Classification/ → classify.py
Goal: Build a content-based recommendation engine matching user skills to career paths.
- Dataset:
raw_skills.csv— 15 job roles × 116 unique skill tags (self-built) - Algorithm: TF-IDF vectorization (penalizes generic skills, rewards specific ones) + Cosine Similarity (magnitude-invariant matching)
- Pipeline: Ingestion → Scoring → Sorting → Filtering (Top-3 output)
- Includes Cold Start detection for unmatched input
- Demonstrates: Logic building, pattern matching, recommendation concepts
📁 Project3_Recommendation_System/ → recommend.py
Goal: Implement both Optical Character Recognition and Object Detection using pre-trained models.
- Path 1 — OCR: Full pre-processing pipeline (Grayscale → Gaussian Blur → Hough-line Deskew → Otsu Adaptive Threshold) feeding into
pytesseract. Achieved 89.6%–93.8% confidence on noisy, skewed test scans. - Path 2 — Object Detection: Transfer learning with pre-trained MobileNet-SSD (Caffe, VOC-trained) via
cv2.dnn. Detected horse (100%), car (99.3%), and person (95.8%) on a multi-object test scene, with an 80% confidence gate correctly filtering out weaker detections. - Demonstrates: Using AI libraries, understanding model outputs, transfer learning
📁 Project4_Image_Text_Recognition/ → ocr_recognition.py, object_detection.py
| Category | Tools |
|---|---|
| Language | Python 3.12 |
| ML / Data | Scikit-Learn, Pandas, NumPy |
| Computer Vision | OpenCV (cv2.dnn), Tesseract OCR (pytesseract) |
| Concepts | Control Flow, Supervised Learning (KNN), TF-IDF + Cosine Similarity, Transfer Learning (MobileNet-SSD) |
Each project folder is self-contained with its own README.md and run instructions. General setup:
git clone https://github.com/KanwalAi/DecodeLabs-AI-Internship-2026.git
cd DecodeLabs-AI-Internship-2026
# Install shared dependencies
pip install scikit-learn pandas numpy opencv-python pytesseract
# System dependency for Project 4 (OCR engine)
sudo apt-get install tesseract-ocr # Linux
# brew install tesseract # Mac
# or download from: https://github.com/UB-Mannheim/tesseract/wiki (Windows)Then cd into any project folder and follow its individual README.md.
This internship moved progressively from deterministic logic → statistical learning → similarity-based matching → perceptual AI:
Project 1 Project 2 Project 3 Project 4
Rule-Based → Supervised → Content-Based → Computer Vision
(if-else) Learning (KNN) Filtering (TF-IDF) (OCR + Detection)
Each milestone builds directly on the last — from teaching a machine to follow explicit rules, to teaching it to recognize patterns in structured data, to matching unstructured preferences, to finally interpreting raw pixels and scanned documents.
The successful completion of this internship is recognized through the following official documents issued by DecodeLabs.
Successfully completed the AI Engineering Internship (Batch 2026), demonstrating practical skills in Artificial Intelligence through four hands-on projects covering rule-based systems, machine learning, recommendation systems, and computer vision.
📄 View Certificate:
AI Engineering Internship Certificate
Received an official Letter of Recommendation from DecodeLabs in recognition of my dedication, technical performance, and successful completion of the internship program.
📄 View Recommendation Letter:
Letter of Recommendation
Kanwal Fatima
BS Artificial Intelligence Student
Passionate about building intelligent systems that combine AI, robotics, computer vision, and embedded technologies.
🔗 LinkedIn: https://www.linkedin.com/in/kanwal-fatima-72a352357/
🐙 GitHub: https://github.com/KanwalAi
DecodeLabs — Your Digital Lab 🌐 www.decodelabs.tech ✉️ decodelabs.tech@gmail.com 📍 Greater Lucknow, India
Completed as part of the DecodeLabs Artificial Intelligence Industrial Training Program, Batch 2026.