AI Engineering Fundamentals: Build Real LLM Apps in Python

Define AI engineering through definitions and real-world data, then set up your development environment, learn to use Jupyter notebooks, and make your first LLM calls with OpenAI API and OLAMA.

Clone the course repository and set up your development environment to start building AI-powered projects. Actively code along, experiment with the examples, and create your own projects to show recruiters.

Set up a Python development environment for building real LLM apps by installing VS Code Python and Jupyter extensions, configuring UV, and creating a project-specific virtual environment.

Create a local .env file and load your OpenAI API key, then make your first LLM call using a configured environment. Keep key private and avoid committing it to git.

Install OLAMA to run a local open-source model, download Gemma 4 with 4 billion parameters, and interact via the OpenAI-compatible Python SDK on localhost.

Define AI engineering as building production-ready apps on top of foundation models, bridging product needs with practical evaluations, costs, latency, and deployment considerations.

Explore why AI engineering is the fastest growing field and how foundation models, APIs, and integration skills drive high demand for AI engineers across production systems.

Scrape current AI engineer job postings. Classify for true AI engineering roles, filter for foundation model applications, and extract core skills using an llm to inform real-world hiring trends.

Scrape AI engineer job postings from LinkedIn and Indeed with the JobSpy scrapeJobs function, fetch descriptions, filter by AI and engineer, deduplicate, and store results in a JSONL file.

Distinguish ai engineering from ml engineering using real job postings and skills like python, langchain, autogen, and semantic kernel, and see how llms turn this into prompt-driven api calls.

Learn to classify job postings with an LLM, using the OpenAI API and a predefined JSON schema to output structured results that identify AI engineering roles.

Compare how the LLM classifications align with manual job classifications for AI engineering, AI/ML engineering, and traditional ML roles, highlighting Lang Chain, vector databases, and production AI enabled applications.

Extract five shared responsibilities of AI engineers from real-world postings by summarizing 41 classified roles, including building AI and ML systems, retrieval workflows, product integration, model lifecycle, and cross-team collaboration.

Extract concrete, data-driven AI engineering skills from real job postings to guide learning and build real LLM apps in Python.

Learn five ai application categories, ranked from most to least common, built on foundation models by analyzing job postings with an llm, and align your portfolio with real-world user needs.

Learn how workflow automation and internal AI agents streamline repeatable business tasks. Explore agent assist, semi-automated, and fully automated use cases in customer service, sales, and operations.

Learn how AI systems read messy documents—emails, PDFs, invoices, and regulatory filings—and extract structured data to automate processing and decision making.

Master retrieval augmented generation (REG) to bridge questions with the right knowledge. Use semantic search to pull relevant sections from internal documents and present answers with source references.

Examine AI-powered decision support in finance, from Experian's credit analytics and decisioning to automate applicant reviews, risk assessment, and proposed interest rates and limits.

Explore content generation at scale for user-facing outputs, including ads, product descriptions, chatbots, and interfaces. Build systems that ensure high output quality, on-brand consistency, and engaging customer experiences.

Explore the top five ai app categories shaping real llm applications, including workflow automation, document processing, knowledge retrieval, decision support, and content generation, with real world examples.

Turn notebooks into a Python project to build a portfolio-friendly AI job agent that collects postings, summarizes roles, ranks matches, and delivers a final report to your email inbox.

Build and deploy a weekly AI engineering job agent in Python on AWS, demonstrating portfolio-ready skills in LLMs, prompt engineering, and workflow automation through job summaries and rankings.

Choose your own portfolio project path—from guided to fully custom—and apply a three-part workflow: a trigger with input, LLM processing steps, and a final output.

learn how to create a new python project from notebooks using uvinit and uvsync, set up a virtual environment, inspect generated files, and run the project in VS Code.

Build a first ai workflow in a new python project that scrapes ai engineering jobs, classifies true ai roles, extracts skills, and generates an HTML report.

Extend the simple Python workflow by enriching the final html report with job and company summaries, salary and location highlights, and a skill-based ranking and decision-support system.

Explore LLM-based semantic skill matching that compares job requirements to a student profile, classifying skills into matched, partial, and no match, with examples like Python and LLM APIs.

Extend the workflow with an LLM to compute match score from the user profile and preferences and rank jobs in HTML report, with application decision, mismatch summary, and recommended action.

Reveals the final HTML report of the extended workflow, detailing scraping, classification, enrichment, skills and overall matching, producing a weekly top AI engineering chops list with apply recommendations.

Integrate Resend to send the HTML report via email by configuring an API key, environment variables, and a recipient in your Python workflow.

Learn to communicate with LLM APIs by distinguishing instructions from input, applying message rules, and mastering prompt engineering to handle multimodal inputs and assess per-request costs.

Explore the model parameter in the OpenAI API to see how the chosen model processes requests. Read in-IDE SDK and web documentation to discover available model identifiers and options.

Learn to choose and pin specific GPT 5.4 variants, including the mini and nano, with date suffixes, and compare frontier models like GPT 5.5 for quality, latency, and cost.

Explain how instructions and input guide a large language model, covering system and user prompts, developer prompts, and the OpenAI responses API with a job advisor example.

Explore prompt engineering by crafting prompts with roles, task descriptions, rules, tone, constraints, and output formats, and learn about one-shot, few-shot, and markdown practices.

Learn how instruction and input prompts differ in authority, and how prioritizing developer prompts prevents conflicts and mitigates prompt injection risks in AI systems.

Explore improved structured output in python by replacing verbose json schemers with a single source of truth for isAIEngineeringRole and reason, avoiding drift.

Refactor your LLM output handling with Pydantic models to replace raw JSON schemas, gaining less boilerplate, a single source of truth, and typed Python objects from the OpenAI SDK.

Explore how to build a multimodal image analysis pipeline that identifies fridge ingredients and suggests healthy recipes using vision-enabled OpenAI models like GPT 5.4 mini in a Python notebook.

Define and populate the pedantic fridge analysis model to identify visible ingredients, with amount and unit as optional unions, and apply grams, milliliters, or pieces for estimates.

Generate three practical recipe ideas from fridge ingredients detected by the vision model GPT 5.4 Mini using an LLM call, including berry yogurt cottage cheese bowl and veggie egg scramble.

Learn to generate images with the Image API using GPT-Image2, sending text prompts and handling Base64 outputs, while comparing image sizes and quality levels and noting generation costs.

Learn how to classify PDFs as invoices and extract vendor, payment date, amount, and currency using an LLM; the workflow converts unstructured invoices into structured data for downstream processing.

Extract text from PDFs locally and convert to markdown to send only structured text to the LLM, reducing token costs and speeding invoice data extraction.

Count input tokens and estimate costs using client.responses.inputTokens to compare pdf and markdown invoice approaches, showing markdown costs about a third of pdf and why this matters.