3-Day Digital Marketing Bootcamp at ACAMIT: Google Sites, Ads, and SEO
Hands-on 3-day digital marketing bootcamp at ACAMIT (Lagankhel) for culinary, hospitality, and IT...
AI Product Engineering: Building Intelligent Applications with Modern AI
Format: 15 Sessions × 2 Hours | 30 Hours Total
Structure: Three Blocks — Foundations, Engineering, Agents & Deployment
By Arjan KC | Digital Marketing Expert & Educator
15
Session Notes
15
Slide Decks
30
Hours
ℹ️
Overview
Course Sessions (3 Blocks)
Block 1 Foundations & The Exocortex
Sessions 1–4 (10 hours). Establish an AI-native development workspace, master structured prompting, build a personal knowledge base (exocortex), and design context for reliable AI ingestion.
Session 1
Session 1: The AI-Native Workspace
Set up a cloud-backed AI development environment. Learn how generative AI democratises software e...
⏱️ 2 hours Session 2Session 2: Prompting & Structured Outputs
Master LLM prompt engineering and structured output techniques. Learn to enforce JSON and Markdow...
⏱️ 2 hours Session 3Session 3: The Personal Exocortex
Build a personal knowledge management system using Obsidian. Learn how to externalise memory into...
⏱️ 2 hours Session 4Session 4: Context Engineering
Design the data that feeds AI systems. Learn context window management, template design, and how ...
⏱️ 2 hoursBlock 2 AI-Assisted Engineering & Integration
Sessions 5–9 (10 hours). Use AI as a coding partner to scaffold applications, integrate cloud AI APIs, generate vector embeddings, and build a complete RAG pipeline.
Session 5
Session 5: AI-Assisted Software Engineering
Use GitHub Copilot as a development partner to navigate, explain, generate, and debug code. Under...
⏱️ 2 hours Session 6Session 6: Rapid Application Prototyping
Scaffold a lightweight user interface using natural language and AI tools. Translate your product...
⏱️ 2 hours Session 7Session 7: API Integration & Cloud Models
Connect your application to a cloud AI API. Learn API authentication, request/response structure,...
⏱️ 2 hours Session 8Session 8: Grounding AI — Embeddings Basics
Move beyond the model training cutoff. Learn how vector embeddings represent meaning, generate em...
⏱️ 2 hours Session 9Session 9: Building the RAG Pipeline
Build a complete Retrieval-Augmented Generation pipeline. Intercept user queries, search your Obs...
⏱️ 2 hoursBlock 3 Agents, Evaluation & Deployment
Sessions 10–15 (10 hours). Engineer agentic workflows with tool calling, process multimodal data, evaluate for hallucination and security risks, and ship the final prototype in the AI Product Lab.
Session 10
Session 10: Introduction to AI Agents
Move from reactive chatbots to autonomous AI workflows. Learn what AI agents are, how they plan a...
⏱️ 2 hours Session 11Session 11: Tool Calling & Actions
Equip AI with tools to interact with external systems. Implement function calling to give AI acce...
⏱️ 2 hours Session 12Session 12: Multimodal AI & Data Pipelines
Process images, PDFs, and structured data with multimodal AI. Build automated data ingestion pipe...
⏱️ 2 hours Session 13Session 13: Evaluation & Debugging
Test your AI system for correctness, relevance, and consistency. Design adversarial inputs to exp...
⏱️ 2 hours Session 14Session 14: Security, Privacy & Responsible AI
Identify and mitigate AI security risks including prompt injection, data privacy leakage, and bia...
⏱️ 2 hours Session 15Session 15: AI Product Lab & Showcase
Final session: complete your AI product, prepare your presentation, and demonstrate your working ...
⏱️ 2 hoursPresentation Slides
Session 1 Slides
Session 1 Slides: The AI-Native Workspace
Slides for Session 1 — cloud AI stack, the engineering partnership model, environment setup, and ...
Session 2 SlidesSession 2 Slides: Prompting & Structured Outputs
Slides for Session 2 — prompt components, zero/few-shot techniques, chain-of-thought, and enforci...
Session 3 SlidesSession 3 Slides: The Personal Exocortex
Slides for Session 3 — exocortex concept, Obsidian setup, atomic notes, Maps of Content, and vaul...
Session 4 SlidesSession 4 Slides: Context Engineering
Slides for Session 4 — context window constraints, context engineering principles, template desig...
Session 5 SlidesSession 5 Slides: AI-Assisted Software Engineering
Slides for Session 5 — AI development partner model, four Copilot workflows, evaluate-refine cycl...
Session 6 SlidesSession 6 Slides: Rapid Application Prototyping
Slides for Session 6 — prototype mindset, architecture-first thinking, minimal stack, UI generati...
Session 7 SlidesSession 7 Slides: API Integration & Cloud Models
Slides for Session 7 — API anatomy, temperature, authentication security, making real Gemini API ...
Session 8 SlidesSession 8 Slides: Grounding AI — Embeddings Basics
Slides for Session 8 — knowledge cutoff problem, vector embeddings, cosine similarity, embedding ...
Session 9 SlidesSession 9 Slides: Building the RAG Pipeline
Slides for Session 9 — RAG pipeline stages, critical system prompt design, retrieval quality eval...
Session 10 SlidesSession 10 Slides: Introduction to AI Agents
Slides for Session 10 — chatbot vs agent distinction, agent loop, types of actions, safeguards, a...
Session 11 SlidesSession 11 Slides: Tool Calling & Actions
Slides for Session 11 — function calling architecture, tool schemas, tool executor, the calling l...
Session 12 SlidesSession 12 Slides: Multimodal AI & Data Pipelines
Slides for Session 12 — multimodal model capabilities, image data extraction, pipeline architectu...
Session 13 SlidesSession 13 Slides: Evaluation & Debugging
Slides for Session 13 — evaluation dimensions, hallucination patterns, adversarial testing catego...
Session 14 SlidesSession 14 Slides: Security, Privacy & Responsible AI
Slides for Session 14 — prompt injection, data privacy risks, bias and fairness, guardrails, and ...
Session 15 SlidesSession 15 Slides: AI Product Lab & Showcase
Slides for Session 15 — showcase structure, live demo guidance, architecture diagram requirements...
Course Overview
About This Module
AI Product Engineering is a 30-hour practical summer school module for students in computing and problem-solving disciplines. It focuses on the engineering layer between AI capabilities and real-world applications — how to turn foundation model power into software that people can actually use.
The module is built on a foundational principle: AI has democratised software development. Modern AI tools allow students to generate, understand, debug, and iterate on software using natural language. This shifts the relevant skills from implementation syntax to architecture, evaluation, security, and product thinking.
THINK → KNOW → BUILD → CONNECT → EQUIP → EVALUATE → SHIP
The Three Blocks
Block 1: Foundations & The Exocortex
Sessions 1–4. Set up a cloud-backed AI development environment, master structured prompting and JSON output control, build a personal knowledge base using Obsidian, and design context templates for reliable AI ingestion. The knowledge base built here becomes the foundation of the RAG pipeline in Block 2.
Block 2: Engineering & Integration
Sessions 5–9. Use GitHub Copilot as an engineering partner to scaffold applications. Build a Flask backend. Integrate a real cloud AI API. Generate vector embeddings from your knowledge base. Build a complete Retrieval-Augmented Generation (RAG) pipeline that answers questions grounded in your curated knowledge.
Block 3: Agents, Evaluation & Deployment
Sessions 10–15. Design multi-step autonomous agent workflows. Implement tool calling so AI can interact with external systems. Process images and documents with multimodal AI. Evaluate for hallucination, relevance, and consistency. Apply security guardrails against prompt injection. Ship a working prototype in the final AI Product Lab.
Learning Outcomes
By the end of the module, students will be able to:
- Explain the role of generative AI and cloud models within modern software systems
- Identify real-world problems where AI can provide meaningful value
- Apply structured prompt and context engineering to control AI outputs
- Use AI-assisted development tools to design, generate, and debug software
- Build personal knowledge systems (an exocortex) to organise information for AI retrieval
- Design basic RAG workflows using cloud APIs
- Integrate lightweight AI agents and tool-calling capabilities into software
- Evaluate AI applications for reliability, hallucination, privacy, and security risks
- Develop and present a functional prototype addressing a defined real-world problem
What Students Build
- ✅ An AI-native development workspace with cloud API access
- ✅ A personal exocortex (knowledge base) in Obsidian
- ✅ A structured prompt library for their application domain
- ✅ A web application with a real cloud AI API backend (Flask + Gemini)
- ✅ A complete RAG pipeline — questions answered from their own knowledge base
- ✅ Tool calling integration for real-time external data access
- ✅ A test suite with documented evaluation results
- ✅ An AI Risk Report covering security, privacy, and fairness
- ✅ A working prototype demonstrated in the AI Product Lab
Technology Stack
Cloud AI Services
- Google AI Studio + Gemini API — inference and embeddings
- GitHub Copilot — AI-assisted code development
Development
- Python + Flask — backend application framework
- HTML + CSS + vanilla JavaScript — frontend interface
- Visual Studio Code — development environment
Knowledge Management
- Obsidian — local Markdown knowledge base
- NumPy — vector similarity computation
No local GPU required. All AI inference runs via cloud APIs on standard student laptops.
Assessment (Non-Credit Summer School)
- Participation and Practical Labs: 20%
- Product Concept and Architecture: 20%
- Functional AI Prototype: 30%
- Evaluation and Responsible AI Report: 10%
- Final Presentation and Showcase: 20%
