Private corporate batches, live online cohorts and 1-on-1 mentoring in feature engineering, supervised and unsupervised modelling, validation and honest evaluation — taught by a practitioner who runs it in production.
20 years across DevOps, SRE and Security · 10,000+ engineers trained · Trained teams at JPMorgan Chase, Verizon, Nokia and the World Bank
Principal DevOps Engineer & Architect
Rajesh teaches Machine Learning as a workflow rather than an algorithm catalogue: split before you look, baseline before you tune, and validate in a way that matches how the model will be used. Sessions concentrate on the things that decide whether a model survives production — target leakage, imbalanced classes, threshold and calibration choices, time-aware cross-validation, and error analysis by slice — with every concept built in code against real datasets and a scikit-learn pipeline that prevents leakage by construction.
Twenty years across DevOps, SRE and Security, in principal and architect roles at PayPay, SoftwareAG, ServiceNow, JDA Software, Intuit, Adobe and others. He has trained engineers at JPMorgan Chase, Verizon, Nokia, the World Bank, VMware, Oracle, Mercedes-Benz and Airbus — more than 10,000 people personally. He teaches what he runs, not what he reads.
You are booked with a named engineer, and that is who turns up. Marketplaces and larger providers rotate whoever is free, so the person who sold you the agenda is rarely the person teaching it.
The same trainer is available for the next engagement, which matters when a team builds on what it learned last time.
Your batch is assigned a named trainer before it starts, and that is who teaches it. See the full faculty.
Engagements are matched on the tool, not the calendar. For Machine Learning that means a trainer who has run it in production — feature engineering, supervised and unsupervised modelling, validation and honest evaluation — rather than whoever is free that week. You are told who is teaching before you commit, and that person is on the discovery call that shapes the agenda.
Where a batch is large enough to need a second trainer, the pairing is declared up front. The lead trainer stays accountable for the syllabus and the assessment either way.
Principal DevOps Engineer & Architect
Twenty years across DevOps, SRE and Security in principal and architect roles at PayPay, SoftwareAG, ServiceNow, JDA Software, Intuit, Adobe, IBM/Emptoris, Ness, MindTree and Accenture. He has trained more than 10,000 engineers personally, at organisations including JPMorgan Chase, Verizon, Nokia, the World Bank, VMware, Oracle, Mercedes-Benz and Airbus. He teaches what he runs, not what he reads.
Teams of 8–30
Custom agenda, your timezone, onsite or online, NDA-friendly.
Request a quoteIndividual engineers
A private instructor and a curriculum built around your goal.
₹99,999Individuals who want peers
Scheduled batch, max 8 to 10 hours of live instruction.
₹34,999A private batch starts with a discovery call. We look at the stack you actually run — the CI system, the cloud, the constraints — and map the agenda onto it, so examples use your topology rather than a generic one.
Delivery is onsite at your premises, live online, or hybrid, scheduled around your release calendar rather than ours. Batches run 8 to 30 engineers.
Every attendee leaves with recordings, slides, lab repositories and a completion certificate. You receive an attendance and assessment report. Invoicing supports PO and GST.
Talk to us about a private Machine Learning batchMachine Learning is the practice of fitting a function to data so it generalises to examples it has never seen. A model is a hypothesis with adjustable parameters, a loss function scores how wrong it currently is, and an optimisation procedure moves the parameters to reduce that loss. Everything else — algorithm choice, feature design, regularisation, validation — exists to control one thing: the gap between performance on data you trained on and performance on data you did not.
The day-to-day work is not algorithm invention. It is a workflow. Frame the task and split the data before looking at it. Explore, clean and engineer features, and defend every step against leakage, which is the most common reason a model that scored well offline collapses in production. Fit a simple baseline first. Then move up through linear models, trees and ensembles, tuning hyperparameters against a validation strategy that matches how the model will actually be used — time-based splits for forecasting, grouped splits when rows share an entity, stratified folds when classes are imbalanced.
The part that separates practitioners from tutorial followers is evaluation. Accuracy is misleading on imbalanced data, a single train-test split is noisy, AUC hides threshold behaviour, and an uncalibrated probability cannot be used for a business decision. Machine Learning done properly means choosing metrics that match the cost of each error type, selecting an operating threshold deliberately, slicing results to find where the model fails, and explaining predictions well enough that someone will sign off on using them. The finished output is a validated, documented, reproducible model artefact — the thing MLOps then takes to production.
Almost every organisation now has data it has been storing for years and pressure to do something predictive with it. The result is a large number of models being built and a much smaller number being trusted, because the modelling was done without validation discipline.
That gap is where the demand sits. Fitting an estimator is a few lines of scikit-learn and any engineer can do it in an afternoon. Knowing whether the resulting number means anything — whether a feature leaked the target, whether the cross-validation respected time order, whether a 0.92 AUC is good on a dataset with 2% positives, whether the probabilities are calibrated enough to threshold — is what organisations actually pay for. It is also what makes the difference between a model that saves money and one that quietly costs it.
The generative wave has not reduced this. Foundation models are strong on language and images and remain beaten by gradient-boosted trees on the tabular business data that drives churn, credit, pricing, demand and fraud decisions. Classical machine learning is where most measurable value still comes from, and the skill has become more valuable, not less, because far fewer people are learning it properly.
What learning from data actually is, and the discipline that has to surround it. Supervised, unsupervised and reinforcement settings; the hypothesis, loss and optimisation triple; generalisation and the bias-variance trade-off; and the split-before-you-look rule that everything else depends on.
Topics: Supervised, unsupervised and reinforcement settings · Hypothesis, loss function and optimisation · Generalisation, overfitting and underfitting · The bias-variance trade-off · Train, validation and test discipline · Assumptions that make learning work: i.i.d. and stationarity · The end-to-end ML workflow and where each step fails · Setting up a reproducible project: environments, seeds, structure
Where most of the improvement in a model actually comes from — and most of the mistakes. Exploratory analysis, missing values, outliers, encoding categoricals, scaling, and features from dates, text and geography. Then leakage in detail, and the pipeline pattern that makes it structurally hard to commit.
Topics: Exploratory data analysis with intent · Missing values: mechanisms and imputation strategies · Outliers, winsorising and robust alternatives · Categorical encoding: one-hot, ordinal, target and frequency · Scaling and normalisation, and which models need it · Features from dates, text and geography · Interaction and aggregate features · Target leakage: how it happens and how to catch it · scikit-learn Pipelines and ColumnTransformer as leakage prevention
The models that should always be tried first, and the concepts they teach that transfer everywhere else. Linear and logistic regression, cost functions and gradient descent, regularisation and what L1 and L2 each do to coefficients, and reading a linear model as an explanation.
Topics: Linear regression: assumptions, fitting and residual analysis · Logistic regression and the log-odds interpretation · Cost functions and gradient descent intuition · Regularisation: ridge, lasso and elastic net · Multicollinearity and unstable coefficients · Polynomial and spline features · Reading coefficients as an explanation, and its limits · Support vector machines and the kernel idea · When a linear baseline is already good enough
The family that still wins on tabular data. Decision trees and how they split, why a single tree overfits, bagging and random forests, then boosting in depth — gradient boosting and the modern implementations, the handful of hyperparameters that matter, and early stopping.
Topics: Decision trees, splitting criteria and pruning · Why single trees overfit and how depth controls it · Bagging and random forests · Gradient boosting: the additive, residual-fitting idea · XGBoost, LightGBM and CatBoost in practice · Learning rate, depth, subsampling and regularisation · Native categorical handling and high-cardinality features · Early stopping against a validation set · Why boosted trees still beat neural networks on tabular data
Finding structure without labels, and being honest about what it means. k-means and its assumptions, hierarchical clustering, density-based methods, choosing the number of clusters, PCA for compression and decorrelation, projection methods for inspection, and unsupervised anomaly detection.
Topics: k-means: assumptions, initialisation and failure cases · Choosing k: elbow, silhouette and business meaning · Hierarchical and density-based clustering with DBSCAN · Evaluating clusters without ground truth · PCA: variance, components and when to use it · t-SNE and UMAP for inspection, and how to misread them · Anomaly detection with isolation forests and one-class methods · Association rules and market-basket analysis · Turning an unsupervised result into an actionable segment
The module that decides whether any earlier number can be trusted. Cross-validation variants matched to data structure, the full metric landscape for classification and regression, imbalance, threshold selection, calibration, and how to tell a real improvement from noise.
Topics: k-fold, stratified, grouped and time-series cross-validation · Nested cross-validation for unbiased model selection · Confusion matrix, precision, recall and F-beta · ROC-AUC vs precision-recall AUC on imbalanced data · Choosing an operating threshold from the cost of each error · Probability calibration: Platt scaling and isotonic regression · Class imbalance: resampling, class weights and why accuracy lies · Regression metrics: MAE, RMSE, MAPE and their pathologies · Is this difference real? Variance, repeats and significance · Error analysis by slice
Finishing the job. Systematic hyperparameter search, then explaining the model to people who must approve it, then packaging it so someone else can reproduce and deploy it. This is the boundary where machine learning work becomes an MLOps input.
Topics: Grid, random and Bayesian hyperparameter search · Search budgets and avoiding validation-set overfitting · Feature importance and why impurity importance misleads · Permutation importance · SHAP values and partial dependence plots · Global vs local explanation for different audiences · Experiment tracking with MLflow · Serialising a pipeline and pinning the environment · Model cards, assumptions and known limitations · The artefact contract handed to MLOps
We rebuild the agenda around the tools you actually run.
Request a custom agendaTake a dataset with a suspiciously good score, locate the leaking feature, rebuild the preprocessing as a fitted pipeline, and report the honest number.
Work one tabular problem from a regularised linear baseline through random forest to a tuned boosted model, recording what each step actually bought.
Build a fraud-style classifier where accuracy is useless — choose PR-AUC, handle imbalance, calibrate the probabilities and set a threshold from a cost matrix.
Rebuild a forecasting model whose cross-validation was leaking the future, using time-series splits and lag features, and compare the two scores.
Segment a customer base, defend the cluster count, profile each segment, and produce a description a marketing owner could use unaided.
Deliver a complete modelling project: framing, leak-free pipeline, tuned model, honest evaluation, SHAP explanation, tracked experiments and a model card.
Billed yearly at ₹9,996
Max 8 to 10 hours of live instruction
Full program, private instructor
8–30 engineers · custom agenda · onsite or online · PO and GST invoicing
Refunds. If we cancel or postpone a cohort, you get a full refund within 15 days. There is no money-back guarantee otherwise.
Terms. Course material remains licensed to the attendee. Read the terms.
Your data. We don't share it with third parties. Privacy policy.
Certificate of completion
My experience with the AIOps training was positive. The course covered important topics in a structured way, and Rajesh Kumar explained the concepts patiently. I found the practical aspects particularly helpful because they made the technical content easier to understand.
I was looking to improve my understanding of AIOps, and this training helped me achieve that goal. Rajesh Kumar explained the subject in a structured and practical manner. The sessions on different AIOps concepts were informative.
Rajesh's experience and knowledge are exceptional and we learnt invaluable practical knowledge which we can apply in our production environment. Incredibly friendly and gave us a fantastic insight both in-depth and at a high level of the Rundeck product.
Great learning experience from a very knowledgeable instructor with well-prepared course notes. The lab exercises on AWS instance work well to learn the hands-on side of the course.
Rajesh is a very good trainer I have experienced in DevSecOps training. The number of contents in different topics he has posted on the DevOpsSchool public website are amazing and user friendly for beginners and experienced professionals.
The trainer (Rajesh) provided very good sessions on SRE profession. Not only hands-on learning on the tools but also SRE mindset.
| What matters | YouTube + blogs | Generic online course | Freelance marketplace | DevOpsSchool |
|---|---|---|---|---|
| Named practitioner | No | Rarely | Varies per booking | Yes — same trainer each time |
| Production experience | Unknown | Unknown | Unverified | 20 years, named employers |
| Custom agenda | No | No | Sometimes | Built from your stack |
| Onsite delivery | No | No | Sometimes | Yes |
| Lab environment | None | Sandbox that expires | Varies | Your own cloud — skill goes with you |
| Assessment | None | Quiz | Rarely | Assignments + capstone per module |
| Per-attendee certificates | No | Sometimes | Rarely | Yes |
| Corporate invoicing | No | Limited | Varies | PO and GST |
| Post-training support | None | Forum, time-limited | None | Lifetime forum access |
Tell us the team, the stack and the timeline. You'll get a straight answer, not a sales sequence.
Talk to an advisorPrefer to call or email?
More ways to reach us on the contact page.