Chooser Page

Tracks

Track pages are the academy's full workflow units. Use them when a topic and a short example are no longer enough and you need a complete run, saved artifacts, and a judgment you can defend. For the ordered ladder, see the Study Plan — this page is a chooser, not a route.

Study Plan First Steps IOAI Surface Back To Home

Pick By Outcome

What Workflow Do You Need?

Tracks are keyed by the workflow move they produce. Choose the track that fixes your current bottleneck, not the one with the most advanced-sounding title.

Ordering Lives Elsewhere

See The Study Plan

If you want "do this one first, then that one," the Study Plan has the phase ladder with exit gates. This page is the inventory.

Pick The Next Track By Outcome

Choose the next track by the workflow you need, not by which title looks most advanced.

• Python, NumPy, Pandas, Visualization — use this if inspection, slicing, plotting, and simple baselines are not yet mechanical
• scikit-learn Validation and Tuning — use this if the split, baseline, tuning, or calibration decisions still feel shaky
• SVM and Advanced Clustering — use this if geometry, kernels, clustering, or manifold views are your current blind spot
• PyTorch Training Recipes — use this if training loops, checkpoints, and optimization control are the main bottleneck
• ResNet, BERT, and Fine-Tuning — use this if representation reuse and fine-tuning choices are the next meaningful step
• Vision and Audio Workflows — use this if you need modality-aware workflow practice after the core route
• Mock Tasks and Timed Workflows — use this if you need better decisions under time pressure
• Imbalanced Triage and Review Budgets — use this if threshold choice, queue policy, or review budgets are the real decision

Advanced Expansion Tracks

Use these once the core tracks are stable or when a specific weakness clearly demands them — they are electives, not prerequisites.

Representation And Transfer

• Optimization, Regularization, and PEFT
• Representation Reuse and Embedding Transfer
• Speech and Audio Encoders

Perception And Structured Outputs

• Detection and Segmentation Workflows
• Text Workflows Beyond Classification
• RAG And Prompting Workflows
• Structured Post-Model Algorithms
• Problem Adaptation and Post-Processing

Advanced Workflow Drills

• Advanced Unsupervised and Manifold Workflows
• Theory Checkpoint Sheets

What A Track Should Produce

Before leaving a track, the learner should have:

• one explicit baseline artifact
• one comparison artifact
• one weak slice or failure-pattern note
• one clear promote, defer, or stop decision
• one exercise response that can be defended without copying the lab