Professional AI Engineering Course

AI Engineering for BFSI

Master LLM fundamentals, prompt engineering, and advanced AI agent design. Build real-world solutions for Banking, Financial Services, and Insurance.

53 Hours Total
Max 15 Students
Weekends Only

Key Learning Outcomes

Understand LLM fundamentals and capabilities
Create effective prompts for various tasks
Build and optimize RAG systems
Learn AI agent architecture and components
Develop stateful AI agents with LangGraph
Work with Multi-Agent Systems and protocols like MCP, A2A
Create AI solutions for BFSI use cases
Deploy agents to cloud platforms
Build evaluation frameworks
Implement monitoring and improvement processes
Integrate Human-in-the-Loop systems

Learn from Industry Expert

RM

Rushikesh Meharwade

Award-winning Data Science Leader • 12+ Years Experience

VP at Motilal Oswal, CEO/CTO at ImmersiLearn. Expert in enterprise-scale Generative AI systems, LLM-powered copilots, and robust MLOps pipelines. Founder of EdTech platform Vidvatta with extensive teaching experience.

Gen AI Initiative 2024
Visionary Leader 2024

10-Week Comprehensive Curriculum

W1🧠LLM Foundations & Basic Prompting
Introduction to Large Language Models and fundamental prompting techniques
6 hours

Module 1: Foundations of Large Language Models (LLMs)

Theory
  • •What are LLMs? Evolution and Key Milestones
  • •Transformer Architecture: A Conceptual Overview (Self-Attention, Embeddings)
  • •Types of LLMs (Foundation Models, Instruction-Tuned Models)
  • •Interacting with LLMs: APIs (e.g., OpenAI, Hugging Face) and SDKs
  • •Understanding Tokens, Context Windows, and Temperature
  • •Ethical Considerations and Responsible AI with LLMs (Bias, Misinformation)
Hands-on
  • •Setting up API access to a chosen LLM provider
  • •Implementing basic API calls for text generation and completion with varying parameters
  • •Analyzing LLM outputs across different temperature settings

Module 2: Advanced Prompt Engineering & LLM Interaction

Theory
  • •Principles of Effective Prompting
  • •Zero-shot, One-shot, and Few-shot Prompts
Hands-on
  • •Designing prompts for summarization, Q&A, and classification tasks (initial exercises)
W2🗣️Advanced Prompting & RAG Introduction
Deep dive into advanced prompting techniques and introduction to RAG systems
6 hours

Module 2: Advanced Prompt Engineering & LLM Interaction

Theory
  • •Instruction Prompting and Role Prompting
  • •Chain-of-Thought (CoT) and Tree-of-Thought (ToT) Prompting (conceptual)
  • •Techniques for reducing hallucinations and improving factual accuracy
  • •Output structuring (e.g., requesting JSON, XML)
  • •Introduction to prompt management and versioning tools
Hands-on
  • •Implementing few-shot prompts for a specific BFSI query type
  • •Experimenting with CoT-style prompts to solve a multi-step problem
  • •Forcing structured output from an LLM

Module 3: Building Retrieval Augmented Generation (RAG) Systems

Theory
  • •What is RAG? Why is it important?
  • •Core components: Vector Databases (e.g., Pinecone, FAISS, Chroma), Embeddings, Retrievers
  • •Data Ingestion and Preprocessing for RAG (Chunking, Metadata)
Hands-on
  • •Setting up a vector database
  • •Ingesting a sample BFSI knowledge base (e.g., policy documents, FAQs)
W3📚RAG Systems & AI Agent Fundamentals
Building complete RAG systems and introduction to AI agent concepts
6 hours

Module 3: Building Retrieval Augmented Generation (RAG) Systems

Theory
  • •Building a basic RAG pipeline (e.g., using LangChain or LlamaIndex)
  • •Advanced RAG: Re-ranking, Query Transformations, Hybrid Search
  • •Evaluating RAG systems (Context Relevance, Answer Faithfulness)
Hands-on
  • •Building a Q&A system over the ingested documents using a RAG pipeline
  • •Experimenting with different chunking strategies and embedding models

Module 4: Introduction to AI Agents

Theory
  • •Defining AI Agents: Autonomy, Reactivity, Pro-activeness, Social Ability
  • •Components of an AI Agent: LLM as the brain, Tools/APIs, Memory, Planning
  • •Agent Architectures: ReAct (Reason + Act), MRKL (Modular Reasoning, Knowledge, and Language) - conceptual overview
Hands-on
  • •Conceptual design of an agent for a simple task (e.g., a weather information agent)
  • •Identifying tools and memory requirements for the designed agent
W4🤖AI Agents & LangGraph Introduction
Understanding AI agents and getting started with LangGraph framework
6 hours

Module 4: Introduction to AI Agents

Theory
  • •Single-agent vs. Multi-agent systems (MAS)
  • •Use cases where agents excel
Hands-on
  • •Comparing and contrasting different agent architectures for a given problem

Module 5: Building Agents with LangGraph

Theory
  • •Introduction to LangGraph: Core concepts (State, Nodes, Edges, Cycles)
  • •Defining agent state and graph structure
  • •Implementing nodes as functions or LangChain Runnables
  • •Conditional edges for dynamic routing and decision-making
Hands-on
  • •Building a simple tool-using agent with LangGraph (e.g., a calculator agent, a web search agent)
  • •Implementing an agent with conditional logic based on LLM output
W5🔗Advanced LangGraph & Multi-Agent Systems
Advanced LangGraph features and multi-agent system architectures
6 hours

Module 5: Building Agents with LangGraph

Theory
  • •Adding tools and tool calling to LangGraph agents
  • •Managing agent memory and conversation history within the graph
  • •Debugging and visualizing LangGraph execution
Hands-on
  • •Creating a multi-step agent that involves a sequence of actions and tool uses
  • •Adding persistent memory to a LangGraph agent

Module 6: Advanced Agent Architectures: Multi-Agent Systems & Contextual Collaboration

Theory
  • •Principles of Multi-Agent Systems (MAS): Cooperation, Competition, Negotiation
  • •Common MAS Architectures (e.g., Hierarchical, Decentralized, Swarm)
  • •Inter-Agent Communication: Protocols, message formats, and content
  • •Context Sharing Mechanisms
  • •Task Decomposition and Allocation in MAS
  • •Introduction to frameworks supporting MAS (e.g., extending LangGraph for MAS, Microsoft AutoGen conceptual overview)
Hands-on
  • •Design exercise: Conceptualize a simple multi-agent system for a BFSI task (e.g., a customer inquiry router agent paired with a specialized product information agent)
  • •Define roles, responsibilities, and necessary information flow (context) between these agents
  • •Discuss challenges in maintaining shared context and potential solutions for the designed system (expanded practical workshop)
W6🛡️Ensuring Responsible & Secure AI Agents
Security, ethics, and responsible AI practices for agent development
6 hours

Module 7: Ensuring Responsible & Secure AI Agents

Theory
  • •Understanding the AI Threat Landscape: Adversarial Attacks (Prompt Injection, Model Evasion, Data Poisoning), Security Vulnerabilities
  • •Defensive Design and Robustness for AI Agents: Input validation, output sanitization, rate limiting
  • •Privacy-Preserving AI: Concepts (Federated Learning, Differential Privacy overview), data minimization, secure data handling
  • •Bias in AI Systems: Sources of bias (data, model, human interaction), fairness definitions and metrics
  • •Bias Detection and Mitigation Techniques
  • •Agent Safety and Alignment: Defining safety protocols, guardrails, content filtering, preventing harmful generation
  • •Introduction to Explainability and Interpretability (LIME, SHAP conceptual overview)
Hands-on
  • •Workshop: Red-teaming a simple agent for vulnerabilities (e.g., prompt injection examples and defense brainstorming)
  • •Practical: Using a tool or framework to identify potential bias in a sample dataset or LLM outputs
  • •Design: Creating a safety and ethics checklist for an agent development project
  • •Discussion: Analyzing case studies of AI security breaches or ethical failures and discussing preventive measures
W7🏦AI Agents in BFSI & Deployment Introduction
BFSI-specific applications and introduction to deployment strategies
6 hours

Module 8: AI Agents in BFSI: Use Cases & Implementation

Theory
  • •Customer Service Automation: Smart chatbots, virtual financial advisors
  • •Fraud Detection and Prevention: Anomaly detection agents, transaction monitoring
  • •Claims Processing Automation: Document analysis, validation, and routing
  • •Personalized Banking & Investment Advice: Recommendation agents
  • •Regulatory Compliance & Reporting: Information extraction and validation agents
  • •Challenges and ethical considerations in BFSI AI agent deployment
Hands-on
  • •Detailed design of an AI agent for one BFSI use case (e.g., an insurance claims pre-processor)
  • •Implementing a core component of the designed agent using LangGraph and RAG
  • •Analyzing potential data sources and APIs required for the chosen BFSI use case

Module 9: Deployment & CI/CD for AI Agents

Theory
  • •Packaging AI Agent applications (Docker introduction)
Hands-on
  • •Overview of Docker concepts and setting up the environment for agent containerization
W8☁️🚀Deployment & CI/CD for AI Agents
Complete deployment pipeline and cloud infrastructure for AI agents
6 hours

Module 9: Deployment & CI/CD for AI Agents

Theory
  • •Deployment options: Serverless Functions (AWS Lambda, Azure Functions, Google Cloud Functions), Container Orchestration (Kubernetes, Amazon ECS, Azure Kubernetes Service)
  • •Managed AI Platforms (Amazon SageMaker, Azure ML, Google Vertex AI)
  • •API Gateway integration for exposing agent endpoints
  • •Security considerations in cloud deployment
  • •Introduction to CI/CD pipelines for AI agent updates (e.g., GitHub Actions, Jenkins)
  • •Infrastructure as Code (IaC) basics (e.g., Terraform, CloudFormation)
Hands-on
  • •Containerizing a LangGraph agent application using Docker
  • •Deploying the containerized agent to a serverless function or a managed container service
  • •Setting up a simple CI/CD pipeline to automate deployment on code changes

Module 10: Evaluation, Monitoring & Iteration of AI Agents

Theory
  • •Defining success metrics for AI agents (task completion, accuracy, user satisfaction, cost)
Hands-on
  • •Discussion and brainstorming of relevant metrics for different agent types
W9📊🧑‍💻Evaluation, Monitoring & Iteration
Comprehensive evaluation frameworks and human-in-the-loop systems
6 hours

Module 10: Evaluation, Monitoring & Iteration of AI Agents

Theory
  • •LLM and RAG evaluation frameworks (e.g., RAGAS, TruLens, LangSmith)
  • •Evaluating agent decision-making and tool usage
  • •Logging agent interactions, decisions, and errors
  • •Monitoring for performance degradation, data drift, and concept drift
  • •Human-in-the-Loop (HITL) Design Patterns: Review and Correction, Escalation
Hands-on
  • •Implementing an evaluation script for a RAG system or an agent's responses
  • •Setting up basic logging for an agent's actions and LLM calls
  • •Designing a HITL workflow for a specific agent task where human oversight is critical (e.g., final approval for a financial transaction flagged by an agent)
  • •Using an LLM evaluation tool to assess the quality of agent responses
W10🏆✨Capstone Project
End-to-end BFSI AI agent development and presentation
6 hours

Module 11: Capstone Project: End-to-End BFSI AI Agent

Theory
  • •Project scoping and requirements gathering for a BFSI AI agent
  • •System design: agent architecture, data sources, tools, evaluation plan
  • •Iterative development and testing
  • •Presentation of the final project, demonstrating its functionality and evaluation results
Hands-on
  • •Students work individually or in small groups: Define a BFSI problem solvable by an AI agent (e.g., an agent to answer queries about different loan products using bank documentation, a basic KYC document checker, an agent to summarize financial news for risk assessment)
  • •Build the agent using LangGraph, RAG, and appropriate tools
  • •Deploy a prototype of the agent
  • •Develop an evaluation plan and test the agent
  • •Incorporate a simple HITL mechanism
  • •Finalize and present project
Theory - Conceptual Learning
Hands-on - Practical Implementation
Total: 60 Hours (10 Weeks × 6 Hours)

Real-World Use Cases

Customer Service Agent

Retrieves data from internal databases to respond to user inquiries

Multi-Agent Market Analytics

Predicts stock prospects by analyzing news, financial reports, and technical patterns

Compliance Automation

Ensures regulatory adherence using Gen AI tools

Course Details & Benefits

Schedule & Duration
  • • Starting from 3rd week of June
  • • Weekends (Saturday and Sunday)
  • • 3 Hours each day
  • • Total Duration: 55 Hours including industry sessions
Pricing & Payment
₹30,000 + GST

Flexible payment options available

  • • Half at start of the course
  • • Half post 30 days
Additional Perks & Benefits
$20 credits to latest Azure Open AI models
Small batch sizes (max 15 students)
Live coding sessions with integrated IDE
Capstone project presentations to BFSI professionals
Weekly one-on-one career guidance
Industry expert guest lectures
Comprehensive job assistance program
1 year access to course content
Course completion certification

Prerequisites

Required

  • Proficiency in Python programming
  • Basic understanding of machine learning concepts
  • Familiarity with using APIs

Recommended

  • Basic knowledge of cloud computing concepts

Ready to Get Started?

Share your details and we'll connect with you to discuss the course.