016 sessions
Production-grade ML
You know how to train a model. This is what separates one that ships.
- Cost-based thresholds
- Calibration
- Leakage audits
- Nested CV
- Optuna
- SHAP
Build A model that would survive review
3 months · you already code
AI Engineering Bootcamp
A conversion course, not an introduction. Fifty-two sessions from your first production-grade model to a deployed agentic system.
- 3 months
- Duration
- 52
- Sessions
- 78
- Contact hours
- 6
- Modules
- 5 + 1
- Projects
Prerequisites — this course does not teach these
- Programming
- Python at working level — functions, classes, virtual environments, packaging.
- Tooling
- Git and GitHub: branches, merges, pull requests. Comfortable in a terminal.
- Data
- SQL joins and aggregates. pandas for loading, cleaning and reshaping.
- Web
- HTTP verbs and status codes, REST, JSON. You have called and ideally built an API.
- ML fundamentals
- Train/test split, overfitting, a confusion matrix. You have trained at least one scikit-learn model.
- Experience
- You have built and maintained something real. This assumes engineering judgement, not just syntax.
Entry is assessed. A short take-home and a fifteen-minute call before a place is confirmed. Admitting someone who does not meet the bar wastes their money and slows the cohort — if you are not there yet, the eight-month diploma is the right route.
The curriculum
Six modules. Each opens with what it will not teach you.
028 sessions
Deep learning
Neural networks, without the magic.
- PyTorch
- Training loops
- Optimizers & schedules
- Mixed precision
- Debugging runs
Build A trained network, from scratch
0310 sessions
Computer Vision
Detection, video, and the edge.
- OpenCV
- CNNs
- Transfer learning
- YOLO
- Segmentation
- Tracking
- ONNX & TensorRT
Build A vision system on live video
047 sessions
NLP and Transformers
Text, embeddings, fine-tuning.
- Tokenization
- Attention
- Hugging Face
- Fine-tuning
- LoRA & QLoRA
Build A fine-tuned language model
0512 sessions
Agentic AI and LLMs
RAG, tools, agents, guardrails.
- Prompting
- Structured output
- Vector stores
- RAG ×3
- LLM evaluation
- Tool calling
- Agents
- Prompt injection
Build An agentic application
069 sessions
Production and capstone
MLOps, LLMOps, and shipping the thing.
- Docker
- Serving
- CI/CD
- Model registry
- Drift
- LLMOps tracing
- Caching & routing
Build Capstone — deployed, documented, defended
Full session-by-session outline, all 52 sessions, with prerequisites for each module. Download the PDF (16 pages).
Next cohort starts soon. Places are limited to 25.
Applications take ten minutes. We reply within three working days, either way — or just message us on WhatsApp.