Year 3 is where you transition from:
“ML Engineer” → “AI Systems Engineer.”
Year 1: Foundations
Year 2: Applied & Engineering Depth
Year 3: Deep Learning + Systems Core
This is the year your profile begins to look serious internationally.
Goal of Year 3:
- Deep learning mastery (not just using libraries)
- Systems-level understanding (OS, memory, networking)
- Multivariable calculus comfort
- Ability to train and debug neural networks properly
- 4–6 substantial deep learning projects
YEAR 3 — Month-by-Month Plan
Structure:
- CS Focus
- Math Focus
- ML / AI Focus
- Project Milestone
MONTH 1 — Operating Systems Fundamentals
CS
- What is an Operating System
- Processes vs Threads
- CPU scheduling
- Memory basics
- Virtual memory concept
- File systems basics
Math
- Multivariable functions
- Partial derivatives
ML
- Review gradient descent deeply
- Derivatives in optimization
Project
Build:
- Multithreaded Python program (concept-level)
- Memory usage analyzer tool
Goal:
Understand what happens under your ML code.
MONTH 2 — Computer Architecture
CS
- CPU basics
- Cache memory
- RAM vs disk
- Instruction execution
- Parallelism basics
Math
- Gradients
- Chain rule deeply understood
ML
- Deriving backpropagation step-by-step
- Implement simple neural network from scratch (no frameworks)
Project
Build:
- Neural network (1 hidden layer) using only NumPy
Goal:
Understand neural networks mathematically, not just by using libraries.
MONTH 3 — Neural Networks Deep Dive
CS
- Code profiling
- Performance optimization basics
- Memory efficiency
Math
- Matrix calculus intuition
- Jacobian (conceptual)
ML
- Activation functions
- Loss functions deeply
- Initialization strategies
- Vanishing/exploding gradients
Project
Build:
- MNIST classifier from scratch
- Experiment with different activations
Goal:
Feel comfortable debugging training failures.
MONTH 4 — Deep Learning Framework Mastery
Choose one:
- PyTorch (recommended)
- TensorFlow
CS
- GPU basics (conceptual)
- Why GPUs matter
Math
- Optimization review
ML
- Tensors deeply understood
- Autograd system
- Training loops
- Model saving/loading
Project
Build:
- Image classifier using CNN
- Custom training loop (not just high-level API)
Goal:
Move from scratch models → real frameworks.
MONTH 5 — Convolutional Neural Networks (CNNs)
CS
- Parallel computing basics
- Data loaders
- Efficient batching
Math
- Convolution intuition
- Feature maps
- Parameter counting
ML
- CNN architecture
- Pooling
- Regularization in deep networks
Project
Build:
- Custom CNN
- Compare architectures
- Document performance
Goal:
Understand image learning systems.
MONTH 6 — Mid-Year Deep Learning Capstone
Build a serious CNN project:
Examples:
- Crop disease classifier (Uganda context)
- Image-based waste sorting
- Document recognition system
Requirements:
- Data preprocessing
- Model tuning
- Evaluation metrics
- Clear documentation
- Clean GitHub repo on GitHub
Checkpoint:
You should now be confident building CNN-based systems.
MONTH 7 — Recurrent Networks & Sequence Modeling
CS
- Intro to networking basics
- How data flows across systems
Math
- Time-series intuition
ML
- RNNs
- LSTM
- GRU
- Sequence modeling basics
Project
Build:
- Text classification model
- Simple time-series predictor
Goal:
Understand sequential data learning.
MONTH 8 — Transformers & Attention
This is where things become modern.
CS
- High-level distributed systems intro
- Why scaling matters
Math
- Attention mechanism mathematically
- Softmax deeply understood
ML
- Attention
- Transformer architecture overview
- Self-attention mechanism
Project
Build:
- Mini transformer (simplified)
- Attention visualization tool
Goal:
Understand modern AI architecture.
MONTH 9 — Large Model Thinking
CS
- Scalability basics
- Distributed training concepts (theoretical)
Math
- Information theory basics:
- Entropy
- Cross-entropy
ML
- Language modeling basics
- Fine-tuning concept
- Transfer learning
Project
Fine-tune a small transformer model.
Goal:
Understand how models are adapted to tasks.
MONTH 10 — Optimization & Debugging Deep Models
CS
- Performance tuning
- Profiling GPU usage
Math
- Advanced optimization review
ML
- Learning rate scheduling
- Adam optimizer deeply
- Gradient clipping
Project
Experiment:
- Train same model with different optimizers
- Document behavior differences
Goal:
Become comfortable diagnosing training issues.
MONTH 11 — ML Systems Engineering
CS
- Model serving basics
- Inference optimization
- Batching requests
- Latency vs throughput
ML
- Model compression basics
- Quantization concept
- Knowledge distillation
Project
Deploy deep learning model as API.
Goal:
Bridge deep learning & production.
MONTH 12 — Year 3 Capstone
Final Project Requirements:
Build a serious AI system:
- Deep learning model
- Clean data pipeline
- API deployment
- Performance optimization
- Documentation
- Architecture diagram
Example:
- Agricultural yield predictor with CNN + tabular model
- NLP classification system for local languages
- Education performance deep model
Document thoroughly.
Share structured summaries weekly on LinkedIn.
End of Year 3 Outcome
If executed seriously:
You will have:
- Strong neural network understanding
- Transformer-level understanding
- Systems awareness (OS, memory, architecture)
- Deployment capability
- Research-paper reading comfort
At this stage, your profile starts looking internationally competitive.
You are no longer “learning ML.”
You are building AI systems.