DEV Community: Md. Mahamudur Rahman

Building Autonomous AI Agents with DeepSeek, LangChain, and AWS Lambda

Md. Mahamudur Rahman — Sat, 22 Mar 2025 21:39:23 +0000

Introduction: The Rise of Agentic Workflows

Autonomous AI agents are reshaping how businesses approach decision-making, automation, and data-driven workflows. These agents, powered by Large Language Models (LLMs) like DeepSeek, are capable of processing complex tasks, retrieving real-time information, and executing decisions autonomously. As this technology becomes more widely accessible, the potential for AI-driven workflows has expanded, making it an exciting time to explore how to leverage these systems in real-world applications.

In this blog, I’ll walk through how to build an agentic AI workflow using DeepSeek, LangChain, AWS Lambda, and AWS Step Functions, showcasing how these tools can be combined to build scalable, intelligent automation systems.

Why DeepSeek for Autonomous AI Agents?

DeepSeek is an open-source LLM optimized for various use cases that require long context lengths (16K tokens), which is ideal for multi-step reasoning, and advanced NLP capabilities for structured decision-making. It is efficient in AI-driven automation, making it a suitable option for building intelligent agents.

DeepSeek’s flexibility and scalability make it a great choice for integrating into workflows that require dynamic information retrieval and autonomous decision-making.

Step 1: Setting Up the Environment

Before we start building, it’s important to get your environment properly set up. Here's how to get started:

Install Required Packages

You’ll need the following packages:

pip install langchain boto3 sagemaker transformers fastapi uvicorn

Initialize DeepSeek Model in SageMaker

For running the model at scale, it’s best to use Amazon SageMaker:

from transformers import AutoModelForCausalLM, AutoTokenizer
import sagemaker

model_name = "deepseek-ai/deepseek-llm-7b"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

session = sagemaker.Session()
role = sagemaker.get_execution_role()

Step 2: Building an AI Agent with LangChain

To create an intelligent agent, LangChain is an ideal framework. It simplifies the process of integrating LLMs with tools and workflows.

from langchain.llms import HuggingFacePipeline
from langchain.agents import initialize_agent, AgentType
from langchain.memory import ConversationBufferMemory

llm = HuggingFacePipeline.from_model(model_name)

memory = ConversationBufferMemory(memory_key="chat_history")

agent = initialize_agent(
    llm=llm,
    tools=[],  # Tools will be added in later steps
    agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
    memory=memory
)

This setup creates an agent that can remember previous interactions and utilize DeepSeek for reasoning.

Step 3: Enhancing the Agent with AWS-Powered Tools

Adding AWS Lambda Integration

You can extend the agent’s capabilities by integrating AWS Lambda. This allows the agent to invoke serverless functions that can interact with external APIs or perform specific tasks.

Example: Stock Market Price Lookup Tool

import boto3, json

lambda_client = boto3.client("lambda")

def stock_price_tool(query):
    response = lambda_client.invoke(
        FunctionName="GetStockPrice",
        Payload=json.dumps({"ticker": query})
    )
    return json.loads(response["Payload"].read())["price"]

agent.add_tool(stock_price_tool, name="Stock Price Checker")

This tool allows the agent to query real-time stock prices and respond accordingly.

Step 4: Automating the Agent Workflow with AWS Step Functions

AWS Step Functions are key for orchestrating complex workflows. They allow you to define a sequence of tasks in a state machine, ensuring that each step in the process is completed successfully.

Example Workflow: Customer Support Automation

You can define a workflow where customer queries are received, processed, and handled by an AI agent:

Step Functions JSON Definition

{
  "StartAt": "Receive Customer Query",
  "States": {
    "Receive Customer Query": {
      "Type": "Task",
      "Resource": "arn:aws:lambda:customer-service-bot",
      "Next": "Classify Intent"
    },
    "Classify Intent": {
      "Type": "Task",
      "Resource": "arn:aws:lambda:classify-intent",
      "Next": "Query DeepSeek"
    },
    "Query DeepSeek": {
      "Type": "Task",
      "Resource": "arn:aws:lambda:deepseek-response",
      "End": true
    }
  }
}

This example demonstrates how to automate customer support workflows by integrating AWS services like Lambda and Step Functions.

Step 5: Deploying the Agent as a Serverless API

To make the agent accessible, you can deploy it as a serverless API using FastAPI and AWS Lambda. This allows you to expose the agent's capabilities through an HTTP interface.

from fastapi import FastAPI
import requests

app = FastAPI()

@app.get("/query")
def query_agent(text: str):
    response = requests.post("https://sagemaker-endpoint-url", json={"inputs": text})
    return response.json()

Deploying with AWS Lambda and API Gateway

Package FastAPI as a Lambda function.
Expose the API via AWS API Gateway.
Enable inference requests from external services.

This setup allows external services or users to interact with your agent via a simple HTTP interface.

Step 6: Future Enhancements

As the AI landscape continues to evolve, there are several areas to explore further:

🔹 Multi-Agent Collaboration – Multiple agents working together to solve complex tasks, particularly in domains like finance and customer support.

🔹 Autonomous Data Gathering – Leveraging tools like Amazon Kendra for intelligent data retrieval and integration into workflows.

🔹 Fine-Tuning for Industry-Specific Applications – DeepSeek can be fine-tuned to perform specialized tasks in different industries, such as healthcare, finance, and legal.

These areas represent the cutting edge of AI-driven automation, and there’s a lot of potential to explore.

Conclusion

By combining DeepSeek, LangChain, AWS Lambda, and AWS Step Functions, you can build powerful autonomous AI agents that can retrieve, reason, and execute complex workflows. These systems are highly scalable, flexible, and capable of handling a wide range of tasks, from customer support automation to data-driven decision-making.

Key Takeaways:

DeepSeek enables long-context AI reasoning – essential for enterprise-grade applications.
LangChain simplifies building decision-making agents, reducing development time.
AWS Lambda and Step Functions provide automation and scalability, ensuring your workflows are robust and efficient.
Serverless deployment ensures your AI-powered services can scale seamlessly, handling large volumes of requests.

This combination of tools is designed to help builders create powerful, scalable AI systems that meet the demands of modern applications.

Next Steps

Deploy your own DeepSeek-powered agent on AWS and experiment with different tools and workflows.
Explore multi-agent collaboration and build workflows where multiple agents solve complex tasks together.
Fine-tune DeepSeek for domain-specific applications to get the most out of its capabilities in your industry.
Integrate more AWS services like Bedrock to take advantage of advanced AI-powered insights in your workflows.

Let’s build the future of AI-driven automation together! 🚀

Demystifying Transformers Architecture in a Simpler Way

Md. Mahamudur Rahman — Sat, 15 Jul 2023 03:20:11 +0000

Introduction

In the world of Natural Language Processing (NLP), the "Attention Is All You Need" research paper introduced an influential architecture known as Transformers. Published by researchers from Google Brain, Google Research and University of Toronto in 2017, this paper presented a groundbreaking method for teaching computers to understand and generate human language. Read transformers paper here. In this blog post, we will break down the key steps involved in the Transformers architecture, making them accessible and engaging for readers of all ages.

Let's Dive In

Step 1: Word Representation
We begin with a sentence or a sequence of words, such as "I love playing soccer." In order to process these words, we represent each one as a number or a vector. Think of it as giving each word a unique code that the computer can understand.

Step 2: Word Table
Next, we arrange these word representations in a table-like structure. Each word gets its own row in the table, and each column represents a different aspect of the word, such as its meaning or position. This organized setup helps the computer keep track of important details about each word.

Step 3: Query, Key, and Value Matrices
Now we introduce three special matrices: Query (Q), Key (K), and Value (V). These matrices are created by multiplying the word table with certain numbers called weight matrices. It's like mixing the word representations together to form these special matrices.

Step 4: Attention Scores
To understand the relationships between words, we calculate "attention scores" for each word. Imagine each word trying to pay attention to other words based on their relevance. We achieve this by multiplying the Query matrix with the transpose (flipped version) of the Key matrix. It's like measuring how much attention one word should give to another.

Step 5: Measures of Similarity
These attention scores act as measures of similarity between words. We want words that are related to have higher attention scores. This helps the computer identify which words are important for understanding the meaning of a sentence.

Step 6: Making Attention Scores User-Friendly
To make the attention scores easier to work with, we use a mathematical function called Softmax. This function ensures that all the attention scores add up to 1 and emphasizes the more important words. It's like adjusting the focus of the spotlight to highlight the most relevant words.

Step 7: Combining Attention Scores and Value Matrix
Here comes the exciting part! We use the attention scores to combine them with the Value matrix. By multiplying the attention scores with the Value matrix, we get what we call the "attention output." It's like taking the important information from each word and putting them together to form a complete understanding.

Step 8: Attention Outputs for Every Word
We repeat this process for every word in the sentence. Each word generates its own attention output, representing the focused and relevant information specific to that word. It's like each word gets its own unique spotlight moment.

Step 9: Utilizing Attention Outputs
Finally, we can use these attention outputs for various purposes, such as translation or summarization. We can also perform additional calculations using the attention outputs to obtain a final result, customized to the task at hand.

Conclusion

In simpler terms, the Transformers architecture introduced in the "Attention Is All You Need" paper involves representing words as numbers, arranging them in a table-like structure, calculating attention scores to understand relationships, and combining attention outputs for a holistic understanding of language. This approach has revolutionized how computers process and generate human language, opening up exciting possibilities in the field of NLP. With Transformers, computers can now grasp the intricacies of language and communicate with us in a more human-like manner. It gets more exciting when we employ the transformers architecture for Computer Vision and many other modalities and tasks. As an example you can read this blog to see how it works.

Embracing the Unique Reality of the Visually Impaired: Exploring AI Integration for Inclusive Experiences

Md. Mahamudur Rahman — Sun, 09 Jul 2023 15:05:26 +0000

Introduction

Imagine perceiving the world without sight. For individuals born with the disability of blindness, their reality is shaped by a unique blend of senses and experiences. In this blog post, we delve into the perception of those without vision and how their understanding of reality differs from the sighted population. Moreover, we explore the potential of AI integration, specifically through transformers, agents, and tools, to create a more inclusive environment and bridge the gap between different realities.

Seeing Beyond Sight

Blindness, the absence or impairment of visual perception, does not imply a complete lack of vision in all cases. Some individuals with blindness may possess residual vision or light perception. However, their primary means of experiencing the world revolve around their other senses - hearing, touch, taste, and smell. By relying on these senses and their cognitive abilities, the visually impaired develop a unique understanding of reality that differs from that of sighted individuals.

The Power of AI Integration

To achieve a generalized version of these different realities, integrating various AI models becomes paramount. One such model is the transformer, which excels in natural language processing tasks and generating human-like text. By leveraging the capabilities of transformers, we can enhance communication, comprehension, and accessibility for individuals with visual impairments.

Transformers as Agents

AI agents equipped with transformer models can play a pivotal role in assisting the visually impaired in perceiving and interacting with their environment. By utilizing computer vision techniques, these agents can analyze images or live video feeds and generate verbal descriptions of objects, people, and activities present in a scene. This auditory feedback provides blind individuals with a deeper understanding of their surroundings and empowers them to make informed decisions based on the information provided.

Empowering Accessibility through Tools

In addition to AI agents, integrating AI tools can significantly enhance accessibility for the visually impaired. Through advancements in text-to-speech conversion and screen reader integration, blind individuals can now access and interact with written information on digital platforms. These tools bridge the gap between text-based content and auditory perception, empowering individuals with visual disabilities to navigate the digital realm seamlessly.

Creating an Inclusive Future

By embracing AI integration, we can actively work towards a more inclusive future. By leveraging transformers, agents, and tools, we can bridge the gap between different realities and empower individuals with visual impairments to interact with the world on their own terms. The integration of AI models enables a shift towards an environment that embraces and supports diverse perspectives, ensuring that individuals with disabilities can navigate the world with greater independence and autonomy.

Use Case: Building an AI-powered Visual Assistance Application on AWS Cloud

To bring the vision outlined above to life, let's consider an example use case where we build an AI-powered Visual Assistance application using AWS cloud services. This application aims to provide real-time object recognition and audio description for the visually impaired, leveraging the power of AI models and AWS infrastructure.

Technical Details

AWS Rekognition
We can utilize Amazon Rekognition, a deep learning-based image and video analysis service provided by AWS. By integrating Rekognition into our application, we can leverage its powerful computer vision capabilities to analyze images or live video feeds in real time. This service can detect and identify objects, people, and activities present in a scene, providing a foundation for generating verbal descriptions.

AWS Lambda
AWS Lambda can be used to build serverless functions that respond to events triggered by the Visual Assistance application. For example, when an image or video feed is uploaded, Lambda can automatically invoke the appropriate function to process the media using Rekognition and generate descriptive audio feedback.

Amazon Polly
To convert text-based information into high-quality speech, we can utilize Amazon Polly, an AWS service that provides text-to-speech functionality. With Polly, we can convert the text-based content found on digital platforms into spoken words, enabling blind individuals to access and interact with written information seamlessly.

Amazon S3
Amazon Simple Storage Service (S3) can be used to store and retrieve media files, such as images or videos, processed by the Visual Assistance application. S3 provides a scalable and durable storage solution that ensures the availability of the processed media for future reference or accessibility purposes.

Amazon API Gateway
To create a secure and scalable API layer for our application, we can utilize Amazon API Gateway. This service enables us to create, deploy, and manage APIs that provide access to the functionality of our Visual Assistance application. It acts as a bridge between the frontend user interface and the backend services, allowing blind users to interact with the application seamlessly.

AWS CloudFront
AWS CloudFront can be used to deliver the application's frontend user interface, ensuring low-latency and high-speed access to the application from different geographic regions. CloudFront caches and serves the static assets of the application, providing an improved user experience for visually impaired users accessing the application from various devices.

Hugging Face's Transformers Library
We can integrate the Transformers library from Hugging Face into our application. This powerful library provides a wide range of pre-trained models for natural language processing tasks, including text classification, text generation, and question answering. By incorporating Transformers, we can enhance the AI agents' ability to process and understand textual information.

Transformers as Agents
In addition to computer vision techniques provided by AWS Rekognition, we can utilize Transformers as AI agents to process and generate natural language descriptions. For example, when the application detects an object in an image or video feed, the Transformers agent can generate a detailed and contextually relevant verbal description of that object. This provides visually impaired individuals with a more comprehensive understanding of their environment.

AI-Driven Text-to-Speech Conversion
To convert the generated textual descriptions into spoken words, we can leverage the text-to-speech capabilities provided by Hugging Face's Transformers and integrate them with Amazon Polly. By combining the power of both services, we can generate natural and expressive audio feedback that accurately represents the verbal descriptions of the visually impaired users' surroundings.

Accessibility Tools
Hugging Face's Transformers library also offers various accessibility tools that can be integrated into our application. For instance, we can utilize their summarization models to generate concise summaries of long articles or web pages, making it easier for blind individuals to access and comprehend textual content on digital platforms.

Continuous Model Training and Improvement
Hugging Face's Transformers library provides resources and tools for fine-tuning and improving pre-trained models. We can leverage these resources to continuously train and refine our AI agents and tools to ensure they deliver accurate and contextually relevant descriptions, summaries, and responses. This ongoing training process enables the application to adapt and improve over time, providing an enhanced user experience for visually impaired individuals.

By incorporating Hugging Face's Transformers Agents and Tools into the AI-powered Visual Assistance application, we can further enhance its capabilities in natural language processing, text generation, and accessibility. This integration allows the application to provide detailed, accurate, and contextually relevant verbal descriptions, summaries, and responses to blind individuals, empowering them to navigate and interact with the world more effectively.

This example use case demonstrates how AI integration and AWS cloud services can be leveraged to create an inclusive environment and empower individuals with visual impairments to interact with the world on their own terms.

Conclusion

The perception of individuals with blindness is unique, shaped by their reliance on non-visual senses. However, through AI integration and the use of transformers, agents, and tools, we can create a more inclusive environment. By empowering the visually impaired with auditory descriptions, enhancing accessibility to digital content, and embracing AI advancements, we can bridge the gap between different realities and ensure equal participation for all. Let us work together to embrace diversity and create a world that celebrates the unique experiences of every individual.

CreatiGenius: Empowering Your Creative Universe

Md. Mahamudur Rahman — Sat, 01 Jul 2023 04:19:17 +0000

Introduction:

CreatiGenius is a dynamic and comprehensive content generation tool that empowers users to unleash their creative genius. With its diverse range of AI-powered agents and tools, CreatiGenius offers an immersive and customizable experience for generating high-quality content across various domains.

Key Features:

StoryCraft: The StoryCraft module of CreatiGenius combines the power of AI language models and storytelling techniques to create captivating narratives, engaging dialogues, and immersive worlds. Users can explore different genres and writing styles, collaborating with the AI to craft unique and compelling stories.

DesignPro: DesignPro is a cutting-edge design platform within CreatiGenius that simplifies the creation of stunning visual content. With an array of customizable templates, intelligent design suggestions, and seamless integration with graphic elements and images, users can effortlessly bring their creative visions to life.

MeloComposer: The MeloComposer component of CreatiGenius taps into the realm of music generation, providing users with the ability to compose original melodies and harmonies. Leveraging advanced AI algorithms and customizable parameters, MeloComposer allows musicians and content creators to generate unique musical pieces that resonate with their desired style and mood.

ArtVisage: ArtVisage, the visionary module of CreatiGenius, merges the world of AI-powered image processing and artistic expression. By utilizing computer vision techniques and generative models, users can generate visually stunning and imaginative artwork, exploring a myriad of styles, from abstract to hyper-realistic.

Infinite Possibilities: CreatiGenius enables users to seamlessly combine and customize content generated by different modules, providing an interconnected creative ecosystem. Users can experiment, iterate, and collaborate with the AI agents, giving birth to limitless creative possibilities.

Motivation:

The motivation behind developing CreatiGenius was to unlock the creative potential within each individual and provide a comprehensive tool that transcends the boundaries of traditional content generation. We aimed to foster a harmonious collaboration between human creativity and AI intelligence, enabling users to express their unique ideas and visions in innovative ways. The project is being built for AWS Community Builders AI/ML Transformer Tools Hackathon in mind.

Process:

The development process of CreatiGenius involved meticulous training and fine-tuning of AI models specialized in various creative domains. We curated extensive datasets, integrated advanced algorithms, and engineered an intuitive user interface to facilitate seamless interaction with the tool's agents and modules. Rigorous testing and user feedback guided us towards refining and enhancing the user experience.

Outcome:

CreatiGenius has emerged as a groundbreaking tool that democratizes the process of content creation. It empowers users to tap into their creative genius, providing them with the tools and resources to bring their ideas to life. With CreatiGenius, users can experience the joy of exploring their creative universe and effortlessly manifest their imaginations into reality.

Reflection:

The journey of creating CreatiGenius has been a testament to the symbiotic relationship between human creativity and AI capabilities. We have witnessed the immense potential of AI-powered content generation in augmenting human creativity, pushing the boundaries of what's possible. This experience has enlightened us on the evolving role of AI in creative domains and inspired us to continue exploring new frontiers.

Conclusion:

CreatiGenius is your gateway to an expansive and limitless creative universe. With its amalgamation of AI intelligence and human ingenuity, CreatiGenius is poised to revolutionize the content generation landscape. Unleash your creative genius, explore uncharted territories, and embark on a journey of boundless creativity with CreatiGenius!

Link to the GitHub Repo: https://github.com/sohagmahamud/CreatiGenius.git