Take a business problem from a one-paragraph ask to a cleaned dataset, a validated model or a grounded LLM assistant, an honest evaluation and a report a decision-maker can act on. About 35 hours of self-paced work over 8 weeks, assessed on AssessAll.
The Applied AI/ML Practitioner course is Level 2 of the Bodhih AI/ML certification, a self-paced online applied machine learning course from Bodhih Training Solutions, Bengaluru. It assumes you can read and modify Python. Over about 35 hours and 8 weeks you work through a diagnostic, eight code-along modules, two labs, an applied project reviewed by a subject-matter expert, a mock and a certification exam. Passing earns a verifiable credential with per-domain readings, re-measured at 90 days. It costs ₹29,999 plus GST.
You followed the tutorial, the accuracy looked good, and you are not sure whether information leaked from the test set, whether the metric suits the problem, or whether the number would survive new data.
A demo that answers three questions is easy. Knowing whether retrieval found the right passage, whether the answer is faithful to it, and where the assistant fails is the part nobody showed you.
Your code reviews cover software. No one has yet read your validation design, your cleaning log or your write-up and told you what a decision-maker would question.
You write and ship code already and want to add tabular ML and LLM-based features to what you build, with evaluation you can stand behind.
You work in pandas or SQL and want to move from describing data to building and validating models on it.
You test or specify systems that now include models, and you need to read a metric, a calibration table or an error analysis and know what it tells you.
Engineering managers and L&D leads who want each person measured per domain before and after, with work reviewed by a subject-matter expert.
If you cannot yet read and modify Python, this course will move too fast for you; AI/ML Foundations (Level 1) is the better place to start, and if you already run ML systems in production, look at the Level 3 AI/ML Specialist instead.
A course records that you watched. This records that your work changed. By the end you hold working artefacts from the labs and the project, each one checked by code, by a subject-matter expert or by both, and a credential that states your reading in each of the seven Level 2 domains.
A loan dataset you have loaded, inspected and cleaned, with missing values, encoding, scaling and dates handled inside a scikit-learn Pipeline and ColumnTransformer so nothing leaks.
Two models built in pipelines for Lab 1, evaluated with a stratified five-fold precision, recall and AUC harness, calibrated, and justified in writing.
A retrieval-augmented assistant over twelve policy documents, scored on recall@3, answer match, faithfulness and latency against eight golden questions, with one failure analysed.
Your Lab 1 model packaged and wrapped in an endpoint, with a working view of batch versus online use, simple monitoring, drift and versioning.
A written record of what your model is for, what data it saw, the bias checks you ran in code and the limits a user should know about.
A six to ten page report with code and a model card on a problem you chose, graded on five criteria by a subject-matter expert after three steering checkpoints.
The Certified Applied AI/ML Practitioner credential, issued on AssessAll and as an Open Badge 2.0, showing your reading in each domain. The diagnostic is re-sat at 90 days to show whether the capability held.
The course runs in 17 stages: a diagnostic, eight code-along modules, two labs, three project checkpoints and the project submission, then a mock and the certification exam. As you read the stage list, notice which of these steps your current work skips.
Measured before you start and again on the same instrument at the end, so the change is recorded rather than claimed.
A 35-minute diagnostic: 28 questions across the seven Level 2 domains, including short pieces of pandas and scikit-learn to read. Not scored for the credential; it sets your baseline per CAP.AI code and is re-sat at 90 days.
pandas idioms, numpy arrays, the scikit-learn estimator API, reading and writing data, reproducibility (seeds, environments). Code-along: load, inspect, clean and summarise the loan dataset in the Pyodide sandbox; notebook download for own environments.
Missing values, encoding, scaling, dates, text basics, leakage, pipelines. Code-along: build a Pipeline with ColumnTransformer on the loan dataset.
Linear and logistic regression, trees, random forests, gradient boosting, model selection, hyper-parameters. Code-along: three models on the ticket-routing dataset.
Cross-validation, stratification, metrics by problem type, calibration, error analysis, learning curves, statistical honesty. Code-along: the stratified 5-fold precision/recall/AUC harness used in Lab 1.
Loan-approval dataset with four planted problems. Clean, build two models in pipelines, evaluate with the M4 harness, calibrate, write a justification. Auto-checks (leakage assertion, AUC floor 0.70, calibration numbers) 50% + SME rubric 50%; pass ≥ 65%. Placed after M4 because M7 wraps this model in an endpoint and M8 writes its model card.
k-means, hierarchical, DBSCAN, PCA, and when clustering misleads. Code-along: segment the customers dataset.
API calls, structured outputs, prompt templates, RAG from scratch (chunk, embed, retrieve, generate), tool use, evaluating LLM outputs with rubrics and LLM-as-judge. Code-along: a 60-line RAG over the policy documents — the seed for Lab 2.
Twelve policy documents, eight golden questions. Chunk, index, retrieve, prompt, generate with a pluggable llm(); score recall@3, answer match, faithfulness and latency; analyse a failure. Auto-checks 50% + SME rubric 50%; pass ≥ 65%.
Packaging a model, a FastAPI endpoint, batch versus online, simple monitoring, drift in one picture, versioning. Code-along: wrap Lab 1's model in an endpoint.
Bias checks with real code, model cards, data-handling records, documentation that survives an audit. Code-along: produce a model card for Lab 1.
Due Day 35. Chosen or own problem, the decision it informs, success measure and proxy metric, first look at the data, risks, two-week plan. SME steers in the comment thread within 3 working days. Required; not marked.
Due Day 42. Data card and cleaning log, validation design, dumb baseline and one real model with the CP1 metric, the thing that is not working. SME steers. Required; not marked.
Due Day 49. Complete draft report and code, model card, changes since CP2, rubric self-check, AI-use declaration draft. SME flags anything that would score 0. Required; not marked.
Due Day 52. 6–10 page report plus code and model card. Rubric-graded: five criteria × 0–4; pass ≥ 14/20 with no criterion at 0; one resubmission. Includes the AI-use declaration and integrity checks.
Fifty questions, seventy-five minutes, including code-reading. Drawn from the Level 2 practice pool; returns per-domain readings and the lessons to revisit before the certification exam.
Sixty questions, ninety minutes, under exam conditions, drawn from the sealed Level 2 pool and including code-reading. Pass: 70% overall and at least 60% in every domain. The credential also needs both labs at 65% or more and the applied project at 70% or more. One free retake after seven days.
Finishing this pathway issues a credential backed by the evidence you produced along the way. It is verifiable by link and is also issued as an Open Badge 2.0, so an employer can check it without taking your word for it.
Handle missing values, categorical encoding, scaling, dates and basic text inside a Pipeline, and learn to spot target leakage before it flatters your score.
Fit linear and logistic regression, trees, random forests and gradient boosting, then choose between them and tune hyper-parameters on a ticket-routing dataset.
Use cross-validation, stratification, metrics by problem type, calibration, learning curves and error analysis to say honestly how good a model is.
Make API calls, get structured outputs, write prompt templates, build RAG from scratch (chunk, embed, retrieve, generate), add tool use and evaluate outputs with rubrics and LLM-as-judge.
Package a model, serve it from a FastAPI endpoint, decide between batch and online scoring, and set up simple monitoring and versioning.
Frame a problem and its success measure, document the data, beat a dumb baseline, and write a report that a non-specialist could act on.
Want the detail before you decide? Read the guide: From Python to applied machine learning: a practical roadmap for developers.
| What is compared | A typical AI course | Applied AI/ML Practitioner at Bodhih |
|---|---|---|
| Starting point | Everyone begins at lesson one, whatever they already know. | A 35-minute diagnostic sets your baseline in each of the seven domains before you start. |
| How you learn | You watch someone else write the code. | Eight code-along modules where you write it, on loan, ticket-routing, customer and policy-document datasets. |
| Practice | Optional exercises with no check on the result. | Two required labs, scored half by automatic checks such as a leakage assertion and half by a subject-matter expert. |
| Project feedback | A project you mark yourself, if there is one. | Three checkpoints where a subject-matter expert steers your work, then a graded submission on a five-criterion rubric. |
| Final assessment | A short quiz or none. | A fifty-question mock, then a sixty-question, ninety-minute certification exam under exam conditions with code-reading. |
| What you leave with | A certificate of completion. | A verifiable credential with per-domain readings, also issued as an Open Badge 2.0, and a re-measurement at 90 days. |
Straight answers on time, price, tools and the credential. If yours is not here, ask us and you will have a reply the same working day.
Yes. The Applied AI/ML Practitioner from Bodhih is an applied machine learning course written for people who can already read and modify Python. It skips programming basics and goes straight to pandas, scikit-learn pipelines, supervised and unsupervised learning, evaluation, building with LLMs, a FastAPI endpoint and responsible practice, with two labs, a reviewed project and a certification exam.
The Applied AI/ML Practitioner course takes about 35 hours of work, planned over 8 weeks (56 days). It is self-paced and online, so you fit the eight modules and two labs around your job. The applied project has three checkpoints set for days 35, 42 and 49 of the plan, with the final submission on day 52, followed by the mock and the certification exam.
The Applied AI/ML Practitioner certification costs ₹29,999 plus GST for an individual, and ₹24,000 per seat for companies. The 18% GST applies to Indian billing addresses. Individuals pay online by UPI, card or netbanking, get access the moment payment is confirmed, and receive a GST invoice. The fee covers all 17 stages, including both labs, the project review and the certification exam.
You need working Python, not advanced maths. The Applied AI/ML Practitioner course assumes you can read and modify Python code, and the diagnostic includes short pieces of pandas and scikit-learn to read. The modules explain each method through code you run yourself. If you do not code yet, the Level 1 course, AI/ML Foundations, is the better starting point.
The Applied AI/ML Practitioner course uses Python with pandas, numpy and scikit-learn for data and modelling, and FastAPI for serving a model. The first code-along runs in an in-browser Pyodide sandbox, with a notebook to download for your own environment. The RAG lab is written around a pluggable llm() function, so the retrieval, prompting and evaluation work is not tied to one model provider.
Yes. One module of the Applied AI/ML Practitioner course is devoted to building with LLMs: API calls, structured outputs, prompt templates, RAG from scratch, tool use and evaluating LLM outputs with rubrics and LLM-as-judge. Lab 2 then has you build a RAG assistant over twelve policy documents and score it on recall@3, answer match, faithfulness and latency.
The Applied AI/ML Practitioner certification is assessed on three things. The exam is sixty questions in ninety minutes under exam conditions, including code-reading, with a pass at 70% overall and at least 60% in every domain. Both labs must reach 65% or more. The applied project must score at least 14 out of 20 on the rubric with no criterion at zero. The exam includes one free retake after seven days.
Yes. The Certified Applied AI/ML Practitioner credential is issued on AssessAll, Bodhih’s sister assessment platform, and can be verified online. It is also issued as an Open Badge 2.0. The credential shows a reading for each of the seven Level 2 domains instead of a single mark, and the diagnostic is re-sat 90 days later to show whether the capability held.
Yes. Companies can buy seats on the Applied AI/ML Practitioner course at ₹24,000 per seat for companies, plus GST, and each engineer receives the same diagnostic, labs, project review and exam. Bodhih also runs in-house workshops for organisations; there is no public price for those, so the team at solutions@bodhih.com or +91 99000 11601 will scope one with you.
The best way to learn applied machine learning as a software engineer is to build complete pieces of work and have them checked. Look for a course that makes you write the code, tests your evaluation and not only your model, includes LLM systems, and has a person review a project. The Applied AI/ML Practitioner does this through eight code-along modules, two scored labs and an expert-reviewed project.
The Applied AI/ML Practitioner is Level 2 of three stacked Bodhih AI/ML certifications. Level 1, AI/ML Foundations, gives a shared vocabulary without requiring code. Level 2 is where coding, labs and the applied project sit, for people who build. Level 3, the AI/ML Specialist in LLM Engineering, goes deeper into LLM system design, agents, evaluation and production concerns.
The Applied AI/ML Practitioner gives you 35 hours of code-along work, two scored labs, a project read by a subject-matter expert and an exam that produces a per-domain reading. It costs ₹29,999 plus GST, or ₹24,000 per seat for companies. Enrol today and start with the diagnostic, so you know where you stand before the first module.
Starting records where you stand today. That baseline is what the re-measurement is compared against, and it is never rewritten afterwards.