AI Lab

Experiments, implementations and systems I'm building while learning AI engineering. What's shipped, what I'm studying and what's planned are kept clearly apart — nothing planned is claimed as known.

See what's built Learning roadmap ↓

3AI features shipped
LLM · CVAreas with real builds
Phase 1Current study focus
PyTorchNext on the plan

The Lab

Built, exploring, next

"Built" links to working code in real projects. "Exploring" and "Next" are honest labels, not credentials.

See the projects →

Roadmap · Planned

Learning roadmap

The sequence I'm following to go from using models to understanding and training them. Only Phase 1 is in progress.

PHASE 01 · NOW · STUDYING

Foundations & Applied Math

Establishing algorithmic rigor, data manipulation fluency, and the foundational mathematics of machine learning.

Python Ecosystem: NumPy vectorized operations, Pandas data munging, Matplotlib/Seaborn analytics.
Mathematics for AI: Linear algebra (matrices, eigenvalues), multivariable calculus (gradients, chain rule), and probability distributions.
Data Structures: Algorithmic problem solving, complexity analysis (Big-O), search and graph traversal.

PHASE 02 · NEXT

Classical Machine Learning

Transitioning from statistical mathematical formulations to predictive modeling and validation pipelines.

Supervised Learning: Linear & logistic regression, decision trees, random forests, SVMs, gradient boosting.
Unsupervised Methods: K-Means clustering, PCA dimensionality reduction, anomaly detection.
Model Evaluation: Cross-validation, precision/recall trade-offs, ROC-AUC, feature selection, regularization.

PHASE 03 · PLANNED

Deep Learning & Transformers

Designing, training, and fine-tuning neural architectures with PyTorch across vision and natural language processing.

Deep Learning Fundamentals: Backpropagation from scratch, activation functions, optimizers (Adam, SGD).
Computer Vision: Convolutional Neural Networks (CNNs), transfer learning, feature representation.
NLP & Transformers: Tokenization, word embeddings, self-attention mechanisms, BERT and GPT architectures.

PHASE 04 · PLANNED

AI Systems & Production Deployment

Packaging AI models into robust, scalable software services with APIs, databases, and monitoring.

GenAI & RAG: Vector databases (Chroma, Qdrant), semantic search, retrieval pipelines, LLM agent tool-calling.
Backend Engineering: FastAPI endpoints, asynchronous processing, PostgreSQL persistence.
MLOps & Shipping: Containerization with Docker, CI/CD pipelines, latency benchmarking, and model evaluation.