Learning Path - Adaptly

Start here

What to do next

Take the placement check below (about 3 minutes).
Waitlist — register for private beta; access is granted manually: /waitlist/
Complete highlighted starter modules if you want extra drills.
Optional: Manual book labs (20 chapters) for deeper system work.

Starter program

Python + ML first steps

Short modules you can finish in one sitting. Recommended ones stay highlighted after placement.

Python: values and flow

12 min · beginner · Write a tiny script that stores values and prints a clear result.

Start here if Python still feels new.

**Do this**
1. Open any Python 3 environment (local, Colab, or Replit).
2. Create variables for a learner name, number of labs done, and total labs.
3. Print one sentence: `Ada completed 2 of 8 starter labs.`

**Check**
- You used at least one number and one string.
- The printed line is readable by a human, not only raw variables.

Practice: Write the print sentence using an f-string or concatenation.

Python: lists and loops

15 min · beginner · Iterate a list and compute a simple summary.

**Do this**
1. Make a list of 5 chapter titles (strings).
2. Loop and print each with its 1-based index.
3. Count how many titles contain the word "data" (case-insensitive).

**Check**
- Indexing starts at 1 in the printed output.
- The count is correct for your list.

Practice: Replace one title and re-run — confirm the count updates.

Python: functions

15 min · beginner · Wrap logic in a function you can reuse.

**Do this**
Write `completion_rate(done, total)` that returns a percentage (0–100).
Call it with (3, 8) and (8, 8).

**Check**
- Division by zero is guarded (return 0 or raise a clear error).
- Results are numbers you can compare.

Practice: Add a second function that returns a short status label: "started", "halfway", or "done".

Python: tiny data clean

18 min · beginner · Clean a messy list of strings before any ML talk.

Given emails like: `["Ada@X.com ", " bob@y.com", "", "Ada@X.com"]`

**Do this**
1. Strip whitespace.
2. Lowercase.
3. Drop empty values.
4. Deduplicate while keeping order.

**Check**
- Final list has unique, normalized emails only.

Practice: Explain in one sentence why cleaning comes before modeling.

ML: problem framing

12 min · beginner · State a learning problem without naming a model.

Pick one real scenario (spam, churn, fraud, recommend next lab).

Write four lines:
1. **User / decision** — who acts on the output?
2. **Input** — what data is available at decision time?
3. **Output** — label or score?
4. **Failure** — what goes wrong if the model is bad?

**Check**
- You did not mention a library yet.
- Failure is concrete (cost, risk, or user harm).

Practice: Circle one input that would leak future information if included.

ML: features vs labels

12 min · beginner · Separate X from y in plain language.

For a "will the learner finish lab 3?" problem:

**Do this**
- List 5 candidate features.
- Name the label.
- Mark one feature that is invalid at prediction time (leakage).

**Check**
- Label is what you want to predict, not an input.
- At least one leakage risk is identified.

Practice: Rewrite one feature so it is available *before* the outcome.

ML: train / validation / test

15 min · intermediate · Explain why splits exist — then apply a simple split rule.

**Do this**
You have 1,000 learner rows sorted by signup date.

1. Propose a time-based split (not random) for train / val / test.
2. Write why random shuffle would be dishonest here.
3. Name one metric you will compute only on the test set at the end.

**Check**
- Test data is never used to pick hyperparameters.
- Split respects time if the product is live.

Practice: Write the split sizes (e.g. 70/15/15) and the date cutoffs in words.

ML: one honest metric

15 min · intermediate · Pick a metric that matches the decision.

Scenario: alert mentors when a learner is stuck.

**Do this**
1. Choose precision, recall, or F1 — and justify in two sentences.
2. Give an example of a false positive and a false negative.
3. Say which error is more expensive for Adaptly beta.

**Check**
- The metric matches the cost of mistakes.
- You can explain the metric without jargon to a teammate.

Practice: Add one segment metric (e.g. new vs returning learners).

Manual companion

20-chapter book labs

All labs

For book readers and advanced practice. Not required to start beta.

Operators vs Architects

System thinking for machine learning

Draw two diagrams for any ML system you have used or built: the model view and the system view. Compare both and identify where you are still operating as an operator rather than an architect.

Open lab

Competitions vs Business

Moving from leaderboard thinking to product outcomes

Take a competition-style ML task and rewrite it as a business system. Define the user, decision, metric, failure mode, and production feedback loop.

Open lab

Feature Engineering

Turning raw data into useful model signal

Choose a dataset and design three feature groups: raw, transformed, and behavioural/contextual. Explain which feature could leak future information and how you would prevent it.

Open lab

Data Is Geometry

Understanding datasets through structure and distance

Run a PCA or UMAP-style analysis on a dataset and explain what the projection reveals, what it hides, and which clusters require additional investigation.

Open lab

Correlation, Causation, and Why Your Model Lies

Leakage, bias, and misleading patterns

Audit a model or dataset for leakage and misleading correlation. Find one feature or segment where the model could look good in training and fail in reality.

Open lab

The Interpretability Problem

Explaining model behaviour without pretending certainty

Prepare a model explanation for a technical reviewer and a non-technical stakeholder. Show what the explanation proves and what it does not prove.

Open lab

Evaluation as a Discipline

Evaluation beyond one headline metric

Build an evaluation scaffold with global metrics, segment metrics, and failure examples. Find where the model performs acceptably overall but poorly for a specific slice.

Open lab

Building Systems, Not Demos

Production readiness and operational thinking

Design a minimal production-ready ML pipeline: feature contracts, model registry, readiness gate, monitoring, and rollback path.

Open lab

The Modern ML Stack

Choosing tools by system responsibility

Assemble a minimal ML stack for one project. Explain what handles tracking, serving, monitoring, data quality, and experiment review.

Open lab

RAG Done Correctly

Retrieval as an evaluated system

Design a minimal RAG pipeline with chunking, embeddings, retrieval, and reranking. Define how you would measure Recall@5 and where the pipeline can fail.

Open lab

Agent Architectures: Hype vs Reality

Safe agent loops and control boundaries

Design a safe agent loop with hard limits, allow-list, logging, and fallback. Compare it with a simple prompt chain for the same task.

Open lab

Fine-Tuning vs Prompting

Cost, quality, and operational trade-offs

Compare a prompted baseline with a hypothetical fine-tuned variant. Define the evaluation protocol, expected quality gain, and ROI crossover point.

Open lab

Quantization and Efficient Inference

Latency, throughput, cost, and quality

Compare FP16, INT8, and INT4/AWQ as deployment choices. Explain the trade-off between speed, memory, and output quality for one use case.

Open lab

Building Your ML Machine

Local, cloud, and hybrid compute choices

Estimate your GPU usage pattern and decide between buying, renting, or hybrid compute. Write a break-even logic for your case.

Open lab

API Dependence vs Computational Sovereignty

Provider risk, routing, and auditability

Design a data classification policy and gateway routing rules for an AI product. Include audit logging and a fallback path for provider changes.

Open lab

The Math You Actually Need

Practical mathematical reasoning for ML engineers

Solve three practical reasoning tasks: tensor shapes in attention, A/B test power, and gradient debugging. Explain the engineering consequence of each result.

Open lab

The Data Science Layer

Data audit before model ambition

Run a data science audit: leakage, class imbalance, temporal split violations, missingness, and baseline sanity checks.

Open lab

Competitive Programming and ML Thinking

Algorithmic discipline inside ML systems

Translate an ML retrieval or ranking task into an algorithmic problem. Compare brute-force search with indexed search by latency and recall.

Open lab

How Big Tech Evaluates Engineers

Structured ML system design communication

Take one ML system design prompt and answer it through six steps: objective, data, model, evaluation, serving, monitoring.

Open lab

What You Build Next

Turning the book into a concrete project path

Choose one project path after the book. Define the first two weeks: objective, dataset, prototype, evaluation, and evidence you will collect.

Open lab

Welcome to Adaptly

What to do next

Quick level check

Python + ML first steps

Python: values and flow

Python: lists and loops

Python: functions

Python: tiny data clean

ML: problem framing

ML: features vs labels

ML: train / validation / test

ML: one honest metric

20-chapter book labs

Operators vs Architects

Competitions vs Business

Feature Engineering

Data Is Geometry

Correlation, Causation, and Why Your Model Lies

The Interpretability Problem

Evaluation as a Discipline

Building Systems, Not Demos

The Modern ML Stack

RAG Done Correctly

Agent Architectures: Hype vs Reality

Fine-Tuning vs Prompting

Quantization and Efficient Inference

Building Your ML Machine

API Dependence vs Computational Sovereignty

The Math You Actually Need

The Data Science Layer

Competitive Programming and ML Thinking

How Big Tech Evaluates Engineers

What You Build Next