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.

Study Plan Beginner Path IOAI Path Back To Home

Default Route

Core Workflow Spine

New learners should do the core tracks in order. That is the fastest way to make data handling, evaluation, and training habits stable.

Competition Route

Decision Quality Under Pressure

IOAI learners should use the competition tracks to practice time pressure, hidden-risk thinking, and operating-point decisions.

Advanced Route

Electives, Not Prerequisites

The expansion tracks are useful only after the earlier workflow is stable or when a specific weakness clearly demands them.

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

Default Core Route

If you are learning AI for the first time, follow this order:

1. Python, NumPy, Pandas, Visualization
2. scikit-learn Validation and Tuning
3. SVM and Advanced Clustering
4. PyTorch Training Recipes
5. ResNet, BERT, and Fine-Tuning

Move on only after each track leaves behind one baseline artifact, one comparison, and one short note about the next move.

Competition Route

If you are preparing for IOAI, use this as the pressure layer after the validation and training basics:

1. Vision and Audio Workflows
2. Mock Tasks and Timed Workflows
3. Imbalanced Triage and Review Budgets

Pair these tracks with Decision Clinics and Solved Questions so the workflow turns into judgment instead of just more code.

Advanced Expansion

Use the expansion layer when you have a specific reason, not when you are avoiding the fundamentals.

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
• 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