Skip to content
L-SYSTEM

intermediate

Applied AI for Big4 IT Professionals

A practical AI course for IT professionals at Big4 and other professional-services firms. Not data-science training: you learn to choose the right AI approach, measure quality, cost and risk, and design a PoC that can honestly say no. The whole course follows one running case. You work in the Digital & IT team of the fictional firm Aldervane Advisory LLP and run two pilots: a Triage PoC that suggests and partly automates routing of ENGAGE Service Desk tickets, and a Policy Assistant, a RAG assistant over internal IT policies that later gets limited agent actions. The data is synthetic but carries real-world defects: leakage in the ticket export, a superseded policy version in the index, a restricted client folder. Topics: framing AI use cases; optimization; probability, base rates and Bayesian updating; classical ML (regression, classification, similarity, triage); enterprise documents; evaluation (precision, recall, overfitting, shadow mode, monitoring); neural networks without magic; LLMs and adaptation options; RAG; AI agents, least privilege and prompt injection; Responsible AI and governance, including the EU AI Act, NIST AI RMF and ISO/IEC 42001 as of September 2026; business cases and go / iterate / stop criteria; a capstone proposal. Every topic is taught as pain, solution and cost. About 190 practice items including calculations, diagrams and illustrations in every unit, and three optional Python labs with ready-made data generators: workload forecasting, ticket classification and a RAG prototype. Course language: Russian.

Path through the course

· 19 units

the order is free · the steps below are the order the author suggests, not a conditionView: map · steps

16

Capstone: AI Solution Proposal

~4 h

by the author’s design, after 13 of these: “AI в Professional Services: от задачи к ценности”, “Optimization: когда AI ищет лучшее решение”, “Probability и неопределённость”, “Bayesian reasoning: обновляем уверенность по evidence”, “Machine Learning Fundamentals”, “Regression и forecasting”, “Classification, similarity и triage”, “Работа с текстом и enterprise documents”, “Model Evaluation, overfitting и business metrics”, “Neural Networks и Deep Learning без магии”, “Generative AI и LLM в enterprise”, “RAG и Enterprise Knowledge”, “AI Agents и workflow automation”, “Responsible AI, privacy, security и governance”, “AI Business Case: от идеи к PoC”

final · in the browser, without the assistant

Exam

closed · enrol on the course

The path in words
  • Stage 1 · AI в Professional Services: от задачи к ценности — ~1.5 h
  • Stage 2 · Optimization: когда AI ищет лучшее решение — by the author’s design, after “AI в Professional Services: от задачи к ценности”; ~1.5 h
  • Stage 2 · Probability и неопределённость — by the author’s design, after “AI в Professional Services: от задачи к ценности”; ~1.5 h
  • Stage 2 · Machine Learning Fundamentals — by the author’s design, after “AI в Professional Services: от задачи к ценности”; ~2 h
  • Stage 3 · Bayesian reasoning: обновляем уверенность по evidence — by the author’s design, after “Probability и неопределённость”; ~1.5 h
  • Stage 3 · Regression и forecasting — by the author’s design, after “Machine Learning Fundamentals”; ~1.5 h
  • Stage 3 · Classification, similarity и triage — by the author’s design, after “Machine Learning Fundamentals”; ~2 h
  • Stage 3 · Работа с текстом и enterprise documents — by the author’s design, after “Machine Learning Fundamentals”; ~2 h
  • Stage 4 · Model Evaluation, overfitting и business metrics — by the author’s design, after “Regression и forecasting”, “Classification, similarity и triage”; ~2 h
  • Stage 4 · Optional Lab: Regression на Python — by the author’s design, after “Regression и forecasting”; optional; ~2 h
  • Stage 4 · Optional Lab: Text Classification на Python — by the author’s design, after “Classification, similarity и triage”; optional; ~2 h
  • Stage 5 · Neural Networks и Deep Learning без магии — by the author’s design, after “Model Evaluation, overfitting и business metrics”; ~1.5 h
  • Stage 6 · Generative AI и LLM в enterprise — by the author’s design, after “Neural Networks и Deep Learning без магии”; ~2.5 h
  • Stage 7 · RAG и Enterprise Knowledge — by the author’s design, after “Generative AI и LLM в enterprise”; ~2.5 h
  • Stage 8 · AI Agents и workflow automation — by the author’s design, after “RAG и Enterprise Knowledge”; ~2 h
  • Stage 8 · Responsible AI, privacy, security и governance — by the author’s design, after “Generative AI и LLM в enterprise”, “RAG и Enterprise Knowledge”; ~2.5 h
  • Stage 8 · Optional Lab: RAG Prototype — by the author’s design, after “RAG и Enterprise Knowledge”; optional; ~3 h
  • Stage 9 · AI Business Case: от идеи к PoC — by the author’s design, after “Model Evaluation, overfitting и business metrics”, “Responsible AI, privacy, security и governance”; ~2 h
  • Stage 10 · Capstone: AI Solution Proposal — by the author’s design, after 13 of these: “AI в Professional Services: от задачи к ценности”, “Optimization: когда AI ищет лучшее решение”, “Probability и неопределённость”, “Bayesian reasoning: обновляем уверенность по evidence”, “Machine Learning Fundamentals”, “Regression и forecasting”, “Classification, similarity и triage”, “Работа с текстом и enterprise documents”, “Model Evaluation, overfitting и business metrics”, “Neural Networks и Deep Learning без магии”, “Generative AI и LLM в enterprise”, “RAG и Enterprise Knowledge”, “AI Agents и workflow automation”, “Responsible AI, privacy, security и governance”, “AI Business Case: от идеи к PoC”; ~4 h

Programme · 19 units

1AI в Professional Services: от задачи к ценности

Отличать AI-задачу от обычной автоматизации и аналитики

~1.5 h
2Optimization: когда AI ищет лучшее решение

Понимать optimization как выбор лучшего действия при ограничениях

~1.5 h
3Probability и неопределённость

Интерпретировать вероятность модели как долю на множестве случаев и понимать calibration

~1.5 h
4Bayesian reasoning: обновляем уверенность по evidence

Понимать prior, likelihood и posterior и считать обновление через шансы и likelihood ratio

~1.5 h
5Machine Learning Fundamentals

Различать supervised (classification, regression), unsupervised и generative постановки

~2 h
6Regression и forecasting

Понимать regression как предсказание числа и forecasting как regression во времени

~1.5 h
7Classification, similarity и triage

Различать binary, multiclass и multilabel classification

~2 h
8Работа с текстом и enterprise documents

Различать extraction, classification, retrieval, summarization и generation и подбирать к каждой свою проверку

~2 h
9Model Evaluation, overfitting и business metrics

Различать train, validation и test и распознавать overfitting и износ test set

~2 h
10Neural Networks и Deep Learning без магии

Понимать слой, параметры, loss и gradient descent и считать число параметров слоя

~1.5 h
11Generative AI и LLM в enterprise

Понимать LLM как вероятностную генеративную модель и объяснять природу hallucination

~2.5 h
12RAG и Enterprise Knowledge

Проследить RAG pipeline от ingestion до grounded answer и назвать, где в нём может сломаться ответ

~2.5 h
13AI Agents и workflow automation

Отличать agentic workflow от chatbot и от детерминированного workflow и выбирать между ними

~2 h
14Responsible AI, privacy, security и governance

Распознавать основные категории AI risk на конкретных системах

~2.5 h
15AI Business Case: от идеи к PoC

Заполнять use case canvas и оценивать data, technical и operational feasibility

~2 h
16Capstone: AI Solution Proposal

Синтезировать business, data, model, architecture и governance decisions в одном proposal

~4 h
17Optional Lab: Regression на Python

Пройти полный цикл простого regression experiment от baseline до отчёта

~2 h
18Optional Lab: Text Classification на Python

Построить baseline text classifier на синтетических тикетах и сравнить его с тривиальной базовой линией

~2 h
19Optional Lab: RAG Prototype

Собрать минимальный retrieval для RAG на безопасной синтетической базе знаний

~3 h