Advanced Certification in Applied Gen AI and Agentic Systems in Oman

Why deep learning?

ANN structure, activation functions

Feedforward and backpropagation

Build ANN using Keras

TensorFlow 2.x and Keras API

Loss functions and optimizers

Overfitting and underfitting

Train a churn model on tabular data

NLP use cases: chatbots, sentiment, classification

Preprocessing: tokenization, stop words, lemmatization

Vectorization: Bag of Words, TF-IDF

Hands-on: Sentiment analysis

Logistic regression and Naive Bayes

Word embeddings introduction

Build: spam or feedback classifier

CNN layers: filters, pooling, convolution

Applications: image classification, document scan

Hands-on: Use pre-built CNN (transfer learning)

Save/load models (Pickle, Joblib, TF)

Create Flask/FastAPI REST APIs

Deploy a model (e.g., churn or sentiment)

Agent, environment, state, action, reward

Exploration vs exploitation

Q-learning (tabular)

Hands-on: Grid-world agent simulation

Self-attention and multi-head attention

BERT vs GPT

Positional encoding

Hands-on: Text generation with transformer

Generator vs Discriminator

DCGAN, CycleGAN, StyleGAN

Hands-on: Synthetic image generation

Denoising Diffusion Probabilistic Models (DDPM)

Diffusion vs GANs

Tools: Stable Diffusion, DALL·E

Hands-on: Generate text-to-image outputs

GPT, Claude, Falcon, Mistral overview

Prompt engineering

Fine-tuning domain-specific LLMs

Hands-on: Legal/Medical chatbot

LangChain architecture

Vector databases (FAISS, Pinecone)

Build RAG pipeline

Hands-on: Doc-aware Q&A chatbot

Vision-Language Models: GPT-4V, CLIP, BLIP

Text-to-image/video/audio generation

Hands-on: Generate media from prompts

Content generation pipelines

Brand asset generation

Tools: Sora, Runway, DALL·E

Deep Q-Networks (DQN)

Policy Gradients

PPO (Proximal Policy Optimization)

Hands-on: Train an autonomous agent

Multi-agent collaboration

LangChain dynamic decision agents

Build a simple autonomous assistant

Task decomposition & planning

Reflex vs learning agents

CrewAI orchestration

Hands-on: Multi-agent research + writing task

LangChain tools: Google Search, Wolfram, Zapier

Agent memory: vector, long-term, summary

Context retention strategies

Hands-on: Web-aware document chatbot

OpenAI, Claude function calling

ReAct prompting strategy

Toolformer concept

Hands-on: Reasoning assistant with tools & APIs

Orchestrating prompt → tools → output

Guardrails, observability

Evaluation metrics: latency, task completion

Hands-on: AI assistant to summarize meetings, update CRM, draft reports

Ethics, safety, bias, hallucinations

Gradio, Streamlit, Hugging Face Spaces

Hands-on: Host your GenAI app publicly

LLM-powered task assistant with RAG

LangGraph + CrewAI agent orchestration

Integration with external tools/APIs (e.g., calendar, CRM, Google Search)

ReAct-based decision flow

Deployed on Streamlit or Hugging Face

Performance evaluation (task completion, responsiveness, hallucination %)

Meeting summarizer + action item generator

Business intelligence research assistant

Autonomous marketing campaign writer

Project planner with goal decomposition