Member-only story

AI Agent: Types (Part-4) Open in app

Vipra Singh · Following

Search
16 min read · Mar 8, 2025

Listen Share More

Discover AI agents, their design, and real-world applications.

Posts in this Series

1. Introduction

2. Build Agent from Scratch

3. AI Agent Frameworks

4. Types of AI Agents (This Post)

5. Workflow vs Agent

6. Agent Architectures

7. Multi-Agent Systems

8. Short-Term and Long-Term Memory

9. Agentic RAG

10. Agentic Guardrails

11. Model Context Protocol (MCP)

12. Evaluation

Table of Contents
· 1. Types of AI Agents
· 1.1 Simple Reflex Agent
∘ 1.1.1 Implementation
∘ 1.1.2 Advantages
∘ 1.1.3 Limitations
· 1.2 Model-based Reflex Agent
∘ 1.2.1 Implementation
∘ 1.2.2 Real-World Example: Amazon Bedrock
∘ 1.2.3 Advantages
∘ 1.2.4 Disadvantages
· 1.3 Goal-based Agents
∘ 1.3.1 Implementation
∘ 1.3.2 Real-World Example: Google Bard
∘ 1.3.3 Advantages
∘ 1.3.4 Disadvantages
· 1.4 Utility-based Agents
∘ 1.4.1 Implementation
∘ 1.4.2 Real-World Example: Anthropic Claude
∘ 1.4.3 Advantages
∘ 1.4.4 Disadvantages
· 1.5 Learning Agents
∘ 1.5.1 Implementation
∘ 1.5.2 Real-World Example: AutoGPT
∘ 1.5.3 Advantages
∘ 1.5.4 Disadvantages
· 1.6 Hierarchical Agents
∘ 1.6.1 Implementation
∘ 1.6.2 Real-World Example: UniPi by Google
∘ 1.6.3 Advantages
∘ 1.6.4 Disadvantages
· 1.7 Multi-Agent Systems
∘ 1.7.1 Key features of multi-agent systems (MAS)
∘ 1.7.2 Real-life use cases of multi-agent systems (MAS)
∘ 1.7.3 Advantages
∘ 1.7.4 Disadvantages
· 2. Transformative Applications Across Industries
· 3. Conclusion

In this blog series, we’ve explored AI agents from foundational concepts to practical development:

Blog 1: Introduction to AI agents and their role in intelligent systems.

Blog 2: Building an AI agent from scratch, covering perception, decision-making, and execution.

Blog 3: AI agent frameworks, comparing tools for efficient development.

What’s Next?

In this blog, we’ll discuss different types of AI agents, their implementation, real-world applications,
advantages, and limitations. From simple reflex agents to multi-agent systems, we’ll explore how these
models power automation, decision-making, and intelligent problem-solving.
Let’s get started!

1. Types of AI Agents
Agents in Artificial Intelligence can be categorized into different types based on how agent’s actions affect
their perceived intelligence and capabilities, such as:

Simple reflex agents

Model-based agents

Goal-based agents

Utility-based agents

Learning agents

Hierarchical agents

Multi-Agent Systems (MAS)

 Image by Author

By understanding the characteristics of each type of agent, it is possible to improve their performance and
generate better actions. Let’s have a detailed overview of the types of AI agents.

1.1 Simple Reflex Agent

The simple reflex agents function only when a certain action or condition takes place. These type of agents
in AI responds based on a predetermined set of rules and do not take past experiences into consideration
when making decisions. Such agents are suited only to execute simple tasks that do not require strategic
thinking.
 Credits: Writesonic

1.1.1 Implementation

class SimpleReflexVacuumAgent:
def __init__(self):
self.location = "A"
self.actions = []
def perceive_and_act(self, current_location, is_dirty):
if is_dirty:
self.actions.append("Suck")
print(f"Cleaned {current_location}")
else:
if current_location == "A":
self.actions.append("MoveRight")
self.location = "B"
else:
self.actions.append("MoveLeft")
self.location = "A"
print(f"Moved to {self.location}")

# Execution
agent = SimpleReflexVacuumAgent()
percepts = [("A", True), ("A", False), ("B", True), ("B", False)]
for loc, dirt in percepts:
agent.perceive_and_act(loc, dirt)
This code implements a direct stimulus-response mechanism where environmental perceptions (room
status) trigger predetermined cleaning behaviors.

Operational Principles: Simple reflex agents function based on condition-action rules, meaning they react
directly to the current input (percept) without considering past experiences. These agents are well-suited
for environments that are:

Fully observable (where the agent has complete information about the state).

Deterministic (where outcomes are predictable based on actions).

Mechanism: The agent follows a predefined set of rules to determine actions.

For example, a vacuum cleaning agent alternates between two locations (A & B), cleaning only when dirt
is detected.

Since it operates purely on present conditions, it does not retain memory of previous states or past
decisions.

Suitable For: These agents are ideal for simple, rule-based tasks where decision-making does not require
historical data. Some common examples include:

Automated customer support: A rule-based chatbot that detects keywords like “password reset” and
provides predefined instructions.

Digital thermostats: Turning on a heater when the temperature falls below a specific threshold.

Industrial automation: Basic IoT devices that respond to direct inputs without complex processing.

1.1.2 Advantages
Easy to design and implement, requiring minimal computational resources

Real-time responses to environmental changes

Highly reliable in situations where the sensors providing input are accurate, and the rules are well
designed

No need for extensive training or sophisticated hardware

1.1.3 Limitations
Here are the limitations of the simple reflex agent:

Adaptability to changes in partially observable environments.

Memory to store previous interactions.

Learning capabilities to improve decision-making over time.

1.2 Model-based Reflex Agent

Extending simple reflex architecture, model-based agents maintain internal representations of world states
to handle partial observability. A model-based reflex performs actions based on a current percept and an
internal state representing the unobservable word. It updates its internal state based on two factors:

How the world evolves independently of the agent

 How does the agent’s action affect the world

A cautionary model-based reflex agent is a variant of a model-based reflex agent that also considers the
possible consequences of its actions before executing them.

Credits: Writesonic

1.2.1 Implementation

class ModelBasedVacuumAgent:
def __init__(self):
self.model = {"A": "Unknown", "B": "Unknown"}
self.location = "A"

def update_model(self, loc, status):

self.model[loc] = "Clean" if status else "Dirty"

def decide_action(self, current_loc, is_dirty):

self.update_model(current_loc, is_dirty)
if is_dirty:
return "Suck"
elif self.model["A"] == "Clean" and self.model["B"] == "Clean":
return "Shutdown"
else:
return "MoveRight" if current_loc == "A" else "MoveLeft"
 # Usage
agent = ModelBasedVacuumAgent()
print(agent.decide_action("A", True)) # Output: Suck

By maintaining a world model and resource states, the agent can make informed decisions despite sensor
limitations.

A model-based reflex agent follows the condition-action rule to determine the appropriate action in a given
situation. However, unlike simple reflex agents, it also maintains an internal state to track environmental
changes and enhance decision-making. This allows the agent to operate effectively even in partially
observable environments.

Model-based reflex agents function in a structured process involving 4 key stages:

1. Sense: The agent perceives the current state of the world using its sensors.

2. Model: It constructs an internal model of the world based on observations.

3. Reason: The agent evaluates its internal model and applies predefined rules or heuristics to determine
the best action.

4. Act: The agent executes the chosen action.

Suitable For: Model-based reflex agents are widely used in various industries, particularly in systems
requiring dynamic adaptation and real-time decision-making:

🔹 Smart Inventory Management: AI-driven systems track inventory levels, analyze purchasing history, and
anticipate demand fluctuations.
🔹 Financial Trading Systems: Automated trading agents maintain market models while reacting to real-
time price changes.
1.2.2 Real-World Example: Amazon Bedrock
One of the most notable examples of a model-based reflex agent is Amazon Bedrock.

💡 What is Amazon Bedrock?

Amazon Bedrock is an AI-powered service that uses foundational models to:
✔️ Simulate operations
✔️ Gain insights from real-world data
✔️ Predict outcomes
✔️ Make informed decisions for optimization and planning
🔹 How Amazon Bedrock Works:
It continuously refines its models using real-time data.

It predicts potential scenarios and optimizes decision-making through simulations.

It adjusts its model parameters dynamically, ensuring adaptability and efficiency.

By relying on model-based reasoning, Amazon Bedrock represents a powerful example of how AI can
enhance decision-making in complex environments.

1.2.3 Advantages
 Enhanced Perception & Memory: The agent maintains an internal model to track environmental
changes over time.

Better Decision-Making: Unlike simple reflex agents, it can handle uncertainty in partially observable
environments by reasoning with historical data.

Optimized Task Execution: By maintaining knowledge of the environment, these agents navigate more
efficiently and anticipate future conditions.

1.2.4 Disadvantages
Building and maintaining models can be computationally expensive

The models may not capture the real-world environment’s complexity very well

Models cannot anticipate all potential situations that may arise

Models need to be updated often to stay current

Models may pose challenges in terms of interpretation and comprehension

1.3 Goal-based Agents

Goal-based agents are AI-driven systems that use environmental information to achieve specific objectives.
Unlike simple reflex or model-based agents, these agents determine the optimal sequence of decisions and
actions needed to reach a desired outcome.

These agents employ search algorithms and heuristics to find the most efficient path to their goals. Due to
their structured decision-making, they are suitable for complex tasks requiring strategic planning and
adaptability.
 Credits: Writesonic

1.3.1 Implementation
Objective-Driven Behavior: Goal-oriented agents select actions that maximize progress toward defined
objectives, employing search and planning algorithms.

Goal Pursuit: The agent assesses current state against target, selecting actions reducing distance to goal.
Complex implementations might integrate graph traversal algorithms.

class GoalBasedAgent:
def __init__(self, target):
self.goal = target
self.actions = []
def path_planning(self, current_state):
# Simplified A* pathfinding logic
if current_state == self.goal:
return "Goal achieved"
return "Move closer" if current_state < self.goal else "Adjust path"

agent = GoalBasedAgent(100)
print(agent.path_planning(75)) # Output: Move closer

Suitable For: Goal-based agents are highly effective for tasks requiring structured planning and
adaptability. Some key applications include:
🔹 Autonomous Vehicles: AI-powered cars navigate routes to safely reach a destination while dynamically
adjusting to obstacles.
🔹 Robotics: Goal-based robots perform complex tasks such as warehouse automation, space exploration,
and industrial operations.
🔹 Computer Vision & NLP: AI models analyze images, videos, and text to extract meaningful insights and
respond accordingly.

1.3.2 Real-World Example: Google Bard

Google Bard: As a goal-based agent, Google Bard aims to provide accurate and high-quality responses to
user queries, selecting actions that help users find relevant information. While it is also a learning agent, its
goal-oriented nature ensures its responses align with user needs.

1.3.3 Advantages
Simple to implement and understand

Efficient for achieving a specific goal

Easy to evaluate performance based on goal completion

It can be combined with other AI techniques to create more advanced agents

Well-suited for well-defined, structured environments

It can be used for various applications, such as robotics, game AI, and autonomous vehicles.

1.3.4 Disadvantages
Limited to a specific goal

Unable to adapt to changing environments

Ineffective for complex tasks that have too many variables

Requires significant domain knowledge to define goals

1.4 Utility-based Agents

Utility-based agents are AI-driven decision-makers that select actions based on a utility function — a
measure of how favorable an outcome is. Unlike rule-based or goal-based agents, these agents evaluate
multiple possible actions and choose the one with the highest expected utility, ensuring more flexible and
adaptive decision-making.

They are particularly useful in complex and uncertain environments where multiple options need to be
compared and optimal choices must be made dynamically.
 Credits: Writesonic

1.4.1 Implementation
Quantitative Decision Making: These agents employ utility functions to evaluate action outcomes,
optimizing for maximum expected value rather than binary goal achievement.

Optimization Logic: The utility function combines multiple factors into a single metric, enabling
comparison of qualitatively different outcomes.

def utility_function(cost, time, risk):

return (0.5 * (1/cost)) + (0.3 * (1/time)) - (0.2 * risk)

actions = [
{"cost": 200, "time": 5, "risk": 0.1},
{"cost": 300, "time": 3, "risk": 0.2}
]
best_action = max(actions, key=lambda x: utility_function(x['cost'], x['time'], x['risk']))
print(f"Optimal action: {best_action}")

To determine the best course of action, a utility-based agent follows a structured process:

1. Modeling the Environment: The agent builds a representation of its surroundings, which can range
from simple to highly complex.
 2. Evaluating Utility: It assesses the expected utility of each possible outcome using a probability
distribution and a predefined utility function.

3. Decision Making: The agent selects the action with the highest expected utility to maximize its success.

4. Continuous Optimization: This process repeats at every decision point, allowing the agent to adapt
dynamically to changing scenarios.

Suitable For: Utility-based agents are widely used in optimization-focused applications, including:

🔹 Resource Allocation: Choosing the most efficient way to distribute computing power, bandwidth, or
funds.
🔹 Scheduling & Planning: Selecting the best time slots for tasks based on constraints.
🔹 Recommender Systems: Recommending best flight tickets by balancing budget and travel time.
🔹 Game AI: Decision-making in strategy-based video games and simulations.
1.4.2 Real-World Example: Anthropic Claude
Anthropic Claude, an AI tool whose goal is to help cardmembers maximize their rewards and benefits from
using cards, is a utility-based agent.

How Claude Works:

It assigns numerical values (utility scores) to different user actions, such as purchasing, bill payments,
and redeeming rewards.

It then compares possible actions in each scenario, weighing trade-offs based on utility values.

It leverages heuristics and AI techniques to simplify and improve decision-making efficiency.

By continuously selecting actions with the highest expected utility, Claude ensures that users receive the
best possible financial benefits.

1.4.3 Advantages
Handles Uncertainty: Can make optimal decisions in dynamic and unpredictable environments.

Comparative Decision-Making: Evaluates multiple choices and picks the best possible action.

Flexible & Adaptive: Adjusts strategies based on changing inputs and outcomes.

1.4.4 Disadvantages
Requires an accurate model of the environment, failing to do so results in decision-making errors

Computationally expensive and requires extensive calculations

Does not consider moral or ethical considerations

Difficult for humans to understand and validate

1.5 Learning Agents

An AI learning agent is a software agent that learns from past experiences and improves its performance
over time. Unlike rule-based systems, these agents start with basic knowledge and adapt automatically
through machine learning techniques.
By continuously analyzing feedback and interactions, learning agents refine their behavior, making them
ideal for dynamic and evolving environments.

The learning agent comprises four main components:

Learning Element: Responsible for learning and making improvements based on experiences from the
environment.

Critic: Evaluates the agent’s performance and provides feedback for improvement.

Performance Element: Executes external actions based on insights from the learning element and
critic.

Problem Generator: Suggests new actions to generate informative experiences that enhance learning.
 Credits: Writesonic

1.5.1 Implementation
Adaptive Intelligence Systems: Learning agents improve performance through experience, typically
employing reinforcement learning frameworks.

Learning Mechanism: The Q-table updates based on received rewards, gradually optimizing action
selection.
 import numpy as np

class QLearningAgent:
def __init__(self, states, actions, alpha=0.1, gamma=0.9):
self.q_table = np.zeros((states, actions))
self.alpha = alpha
self.gamma = gamma
def learn(self, state, action, reward, next_state):
max_future_q = np.max(self.q_table[next_state])
current_q = self.q_table[state, action]
new_q = (1 - self.alpha) * current_q + self.alpha * (reward + self.gamma * max_future_q)
self.q_table[state, action] = new_q

# Initialize agent with 5 states and 4 actions

agent = QLearningAgent(5, 4)
agent.learn(1, 2, 10, 3)

AI learning agents operate in a continuous feedback cycle, allowing them to observe, learn, and adapt:

1. Observation: The agent gathers data from its environment using sensors or input sources.

2. Learning: It analyzes data using algorithms and statistical models, identifying patterns and refining its
knowledge.

3. Action: Based on learned insights, the agent decides and acts within its environment.

4. Feedback: The agent receives rewards, penalties, or environmental cues to assess its performance.

5. Adaptation: Using feedback, the agent updates its knowledge and decision-making processes,
continuously improving.

This cycle repeats over time, allowing the agent to refine its decision-making and adjust to changing
circumstances.

Suitable For: Learning agents are particularly useful in scenarios requiring continuous improvement and
personalization:

🔹 E-commerce Personalization: AI-driven recommendation systems analyze user behavior to improve

targeted advertising.
1.5.2 Real-World Example: AutoGPT
A good example of a learning agent is AutoGPT, developed by Significant Gravitas.

How AutoGPT Works:

Suppose a user wants to purchase a smartphone and asks AutoGPT to conduct market research on the
top ten smartphones.

AutoGPT analyzes product features, reviews, and specifications across various sources.

It verifies website credibility using a sub-agent program to ensure reliability.

Finally, it generates a comprehensive report listing the pros and cons of the top ten smartphone
brands.
By observing, analyzing, and refining its process, AutoGPT continuously enhances its performance,
making it a powerful example of a learning agent in action.

1.5.3 Advantages
The agent can convert ideas into action based on AI decisions

Learning intelligent agents can follow basic commands, like spoken instructions, to perform tasks

Unlike classic agents that perform predefined actions, learning agents can evolve with time

AI agents consider utility measurements, making them more realistic

1.5.4 Disadvantages
Prone to biased or incorrect decision-making

High development and maintenance costs

Requires significant computing resources

Dependence on large amounts of data

Lack of human-like intuition and creativity

1.6 Hierarchical Agents

Hierarchical agents are AI systems structured in a hierarchy, where high-level agents oversee lower-level
agents. This structure allows for efficient task management, ensuring that complex goals are broken down
into manageable sub-tasks. The number of levels in the hierarchy depends on the complexity of the
system.

These agents are widely used in domains that require coordination and prioritization of multiple tasks,
such as robotics, manufacturing, and transportation.

Credits: Writesonic

1.6.1 Implementation
Layered Architecture: Hierarchical agents employ multiple abstraction levels, with higher layers handling
strategic decisions and lower layers managing tactical execution.

Delegation Logic: The supervisor agent manages subsystem coordination, demonstrating distributed
responsibility architecture.

class SupervisorAgent:
def __init__(self):
self.subagents = {
"security": SecurityAgent(),
"climate": ClimateAgent()
}
def coordinate(self, sensor_data):
if sensor_data["intruder"]:
self.subagents["security"].activate()
else:
self.subagents["climate"].adjust(sensor_data["temp"])
class SecurityAgent:
def activate(self):
print("Security protocols engaged")
class ClimateAgent:
def adjust(self, temp):
action = "Cool" if temp > 72 else "Heat"
print(f"Climate system: {action} activated")

# System execution
smart_home = SupervisorAgent()
smart_home.coordinate({"intruder": True, "temp": 68})

The operation of hierarchical agents is similar to a corporate organization, where tasks are organized and
managed at different levels:

1. High-Level Agents: Define broad objectives, plan strategies, and break down tasks into smaller sub-
goals.

2. Intermediate-Level Agents (if applicable): Act as coordinators, ensuring efficient task delegation
between high and low-level agents.

3. Low-Level Agents: Execute the assigned sub-tasks and report progress to higher-level agents.

This structured approach ensures efficient execution, resource optimization, and scalability in complex
environments.

Suitable For: Hierarchical agents are ideal for large-scale businesses and enterprises that require
structured task execution. Some key applications include:

🔹 Industrial Automation: Managing manufacturing workflows with multiple production stages.

🔹 Autonomous Robotics: Coordinating high-level navigation and low-level motor control.
🔹 Transportation Systems: Optimizing traffic control and logistics.
1.6.2 Real-World Example: UniPi by Google
UniPi, by Google, is an innovative hierarchical AI agent that utilizes text and video as a universal interface,
enabling it to learn diverse tasks across various environments.
How UniPi Works:

High-Level Policy: Generates instructions and demonstrations based on diverse inputs, including text
and video.

Low-Level Policy: Executes tasks by learning through imitation and reinforcement learning.

The high-level policy adapts to various environments, while the low-level policy refines execution
strategies based on feedback.

By combining high-level reasoning with low-level execution, UniPi demonstrates the power of hierarchical
AI in handling complex, multi-step tasks efficiently.

1.6.3 Advantages
Scalable Task Management: Enables handling of multiple interdependent tasks within large-scale
systems.

Improved Efficiency: Decomposes complex problems into manageable sub-tasks, enhancing execution
speed.

Adaptability: The hierarchical structure allows for dynamic task prioritization and coordination.
1.6.4 Disadvantages
Complexity arises when using hierarchies for problem-solving.

Fixed hierarchies limit adaptability in changing or uncertain environments, hindering the agent’s ability
to adjust or find alternatives.

Hierarchical agents follow a top-down control flow, which can cause bottlenecks and delays even if
lower-level tasks are ready.

Hierarchies may lack reusability across different problem domains, requiring the time-consuming and
expertise-dependent creation of new hierarchies for each domain.

Training hierarchical agents is challenging due to the need for labeled training data and careful
algorithmic design. Applying standard machine learning techniques to improve performance becomes
difficult due to the complexity involved.

1.7 Multi-Agent Systems

Multi-agent systems (MAS) aren’t exactly a different type of AI agent. Rather, they are a collection of agents
that coordinate with each other and function as a single unit.
 Credits: Writesonic

This can comprise single reflex agents, goal-based agents, or any other type of agent mentioned above.

Unlike hierarchical agents, each agent in the MAS has its own goals and capabilities but interacts with other
agents to achieve a common objective or optimize individual outcomes.
1.7.1 Key features of multi-agent systems (MAS)
Decentralization: Decision-making is distributed among multiple agents.

Collaboration and competition: Agents work together or compete, depending on the scenario.

Scalability: They can handle large-scale problems by distributing the workload.

Specialization: Individual agents can focus on specific tasks within the system.

1.7.2 Real-life use cases of multi-agent systems (MAS)

Multi-agent systems are quite complex and find applications in large-scale environments.

A good example is a healthcare AI system that uses multiple agents for patient care coordination, hospital
resource optimization, and medicine delivery. Individually, each of these agents plays an important role
that’s also viable independently.

But, in a system, these agents can help manage entire hospitals of hundreds or thousands of people.

1.7.3 Advantages
 Scalable for complex, large-scale applications.

Offers redundancy and robustness, as tasks can continue even if one agent fails.
1.7.4 Disadvantages
Coordination between agents can be complex.

Potential for conflicts if agents have competing goals.

2. Transformative Applications Across Industries

Credits: Daffodil

Now that you have a better idea of different types of AI Agents, let’s see how businesses can use their full
potential by integrating such autonomous agents across industries. The following table shows industry-wise
use cases of AI agents, transforming the way businesses function and serve their customers.
 Credits: bacancytechnology

3. Conclusion
The taxonomy of AI agents presents a continuum from reactive architectures to sophisticated learning
systems, each finding unique applications across industry verticals. As demonstrated through healthcare
coordination networks and autonomous manufacturing systems, the strategic combination of agent types
within multi-agent frameworks enables solutions to problems of unprecedented complexity. The provided
code examples and architectural patterns offer concrete implementation blueprints while highlighting the
importance of proper system design. Future advancements in neural-symbolic integration and quantum-
enhanced optimization will further expand the capabilities of intelligent agent systems, driving innovation
across all sectors of the global economy.

Credits
In this blog post, we have compiled information from various sources, including research papers, technical
blogs, official documentations, YouTube videos, and more. Each source has been appropriately credited
beneath the corresponding images, with source links provided.

Thank you for reading!

If this guide has enhanced your understanding of Python and Machine Learning:
 Please show your support with a clap 👏 or several claps!
Your claps help me create more valuable content for our vibrant Python or ML community.

Feel free to share this guide with fellow Python or AI / ML enthusiasts.

Your feedback is invaluable — it inspires and guides our future posts.

Connect with me!

Vipra

Agents Agentic Ai Llm Genai Application

Following

Written by Vipra Singh

6.3K Followers · 12 Following

Responses (5)

Next

What are your thoughts?

Med Dhif
Mar 8

Thank you. I have been waiting for part 4 and the rest.

4 1 reply Reply

Ibiloye Abiodun Christian

4 days ago (edited)

Thanks for the educative post, going through the ToC, priceless ! Could not afford the membership, as dollars to naira is unfairly high and
presently recovering..yet involved in academics

1 Reply
 Naynesh Shah
6 days ago

Another detailed blog, thanks Vipra!!

Reply

See all responses

More from Vipra Singh

Vipra Singh

LLM Architectures Explained: Transformers (Part 6)

Deep Dive into the architecture & building real-world applications leveraging NLP Models starting from RNN to
Transformer.

Nov 10, 2024 562 6

 Vipra Singh

LLM Architectures Explained: Word Embeddings (Part 2)

Deep Dive into the architecture & building real-world applications leveraging NLP Models starting from RNN to
Transformer.

Aug 18, 2024 754 9

Vipra Singh

LLM Architectures Explained: BERT (Part 8)

Deep Dive into the architecture & building real-world applications leveraging NLP Models starting from RNN to
Transformer.

Nov 17, 2024 243 2

 Vipra Singh

Building LLM Applications: Introduction (Part 1)

Learn Large Language Models ( LLM ) through the lens of a Retrieval Augmented Generation ( RAG ) Application.

Jan 9, 2024 1.5K 6

See all from Vipra Singh

Recommended from Medium

 In Data Science Collective by Paolo Perrone

The Complete Guide to Building Your First AI Agent (It’s Easier Than You Think)
Three months into building my first commercial AI agent, everything collapsed during the client demo.

6d ago 1.4K 36

Thack

Claude 3.7 Sonnet: the first AI model that understands your entire codebase
Context is king. Emperor Claude is here. In this exhaustive guide to our newest frontier model, we’ll show you exactly how
to make it work.

Feb 25 986 33
 In Level Up Coding by Fareed Khan

Testing 18 RAG Techniques to Find the Best

crag, HyDE, fusion and more!

6d ago 702 11

In Towards AI by DarkBones

You’re Doing RAG Wrong: How to Fix Retrieval-Augmented Generation for Local LLMs
How To Set Up RAG Locally, Avoid Common Issues, and Improve RAG Retrieval Accuracy.

Mar 8 312 8
 In Artificial Intelligence in Plain English by BavalpreetSinghh

Agentic RAG: How Autonomous AI Agents Are Transforming Industry

Introduction

Mar 4 409 7

Irina Adamchic

Build your hybrid-Graph for RAG & GraphRAG applications using the power of NLP
Build a graph for a hybrid RAG/GraphRAG applications for a price of a chocolate bar!

Mar 5 352 3
See more recommendations