Umang Chaudhary

AI/ML Engineer

I thrive in collaborative environments where I can contribute to ambitious projects that push the boundaries of what's possible with artificial intelligence. My work philosophy centers on continuous learning, meticulous attention to detail, and a commitment to delivering excellence in every aspect of AI system development.

AI/ML Engineer with over 3.7 years of experience in designing and deploying cutting-edge AI solutions across Generative AI, LLMs, machine learning algorithms, and deep learning. Skilled in custom model development, fine-tuning LLMs for domain-specific tasks, feature engineering, data preprocessing, EDA, and model optimization for high-performing predictive systems. Proficient in building AI agent workflows, data processing pipelines, and end-to-end deployments using Flask, FastAPI, Docker, and Kubernetes. Experienced with AWS, Azure AI Studio, PostgreSQL, and MySQL, ensuring scalable, secure, and reliable solutions. Strong foundation in data structures, OOPS, and modeling concepts, with a track record of driving automation, efficiency, and business impact through AI-driven systems.

Education

B.Sc Computer Science

Focused on core computer science concepts including Data Structures, Operating Systems, Networking, System Design, and Database Management.

M.Sc Computer Science

Specialized in Machine Learning and Artificial Intelligence. Gained expertise in building intelligent systems, applying statistical models, and implementing advanced AI algorithms for real-world applications.

Blog

Building Scalable AI Pipelines

Explore the best practices for designing and deploying scalable machine learning pipelines in production environments...

AI MLOps Python August 2025

Building Scalable AI Pipelines

AI MLOps Python August 2025
Diagram of a scalable AI pipeline architecture
Scalable AI pipeline architecture
Screenshot of pipeline monitoring dashboard
Pipeline monitoring dashboard

Introduction

Scalable AI pipelines are critical for deploying machine learning models in production. This post explores key strategies for building robust and efficient pipelines that handle large-scale data and ensure reliability.

Key Components

• Data Ingestion: Efficiently handling diverse data sources
• Model Training: Automating training with distributed systems
• Deployment: Using Kubernetes for scalable model serving
• Monitoring: Real-time performance tracking and drift detection

Implementation Details

The pipeline leverages Python, TensorFlow, and Kubernetes to process data in real-time. We use Apache Airflow for orchestration and MLflow for experiment tracking, ensuring seamless integration and scalability.

Conclusion

Building scalable AI pipelines requires careful planning and robust tools. By following these best practices, you can deploy AI solutions that scale efficiently and deliver consistent performance.

Cartio : E-Commerce Based Conservational Chatbot

Fine-tuned Llama3-8B model on conversational dataset using LoRA technique on e-commerce chat based dataset.

Python LLM Fine-tune Llama 3 - 8B LoRA

Cartio : Fine-tuning LLaMA 3 for an E-commerce Conversational Chatbot

Python LLM Fine-tune Llama 3 - 8B LoRA

Project Overview

  • In an e-commerce setting, a chatbot should not just answer queries but also engage in interactive dialogue — asking clarifying questions, guiding the user through product searches, and handling follow-ups effectively.
  • When we tested the standard LLaMA 3 model with prompt engineering, it often confused user and assistant roles, generated irrelevant responses, and skipped essential follow-up questions. For example, instead of asking “What product are you looking for?”, it would directly generate an answer or hallucinate a response. This behavior made the chatbot unsuitable for a real-world e-commerce use case.

Approach to the Solution

  1. Requirement Analysis: Identified the gap between generic model outputs and the specific needs of e-commerce conversations (e.g., product recommendations, order tracking, returns, etc.).
  2. Dataset Preparation: Curated and structured a domain-specific customer support dataset containing real-world e-commerce queries and responses. This dataset emphasized follow-up questions, context retention, and appropriate responses.
  3. Fine-tuning Strategy:
    • Chose LoRA (Low-Rank Adaptation) for parameter-efficient fine-tuning.
    • Trained the model on NVIDIA RTX 4090 GPU, ensuring faster experimentation and optimization.
      Document processing interface showing structured data extraction from various document types with highlighted annotations
      Nvidia RTX 4090 GPU Details
    • Adjusted hyperparameters to balance response accuracy and conversational flow.
  4. Training: Fine-tuned LLaMA 3 - 8B on the prepared dataset until the model learned to correctly distinguish between user queries and assistant roles.
  5. Evaluation: Compared the before vs. after performance by testing with real e-commerce queries.

Results

  • The fine-tuned chatbot demonstrated a significant improvement in conversation quality.
  • Base model response issue, it just focuses on the generating next token without focusing on the user question and eventually ended up doing user and assistant conversation both from it side.
    • Document processing interface showing structured data extraction from various document types with highlighted annotations
    Base model Response for Question : I want to cancel my order #12345. Can you help me?
  • After fine-running with e-commerce based conversational dataset it was able to process and understand correctly and even asking correctly follow up question.
    Document processing interface showing structured data extraction from various document types with highlighted annotations
    Fine tuned model response for same Question : I want to cancel my order #12345. Can you help me?
  • It maintained context over multiple turns, improving the user experience.
  • Responses became more accurate, natural, and aligned with e-commerce workflows like browsing, order status checks, and returns.

Impact

  • Enabled a customer-friendly chatbot that can handle complex conversations without human intervention.
  • Improved user satisfaction through interactive, context-aware dialogue.
  • Reduced support workload by automating common queries like order status and returns.
  • Showcased how domain-specific fine-tuning can transform a general LLM into a business-ready solution.

Get In Touch

I'm always interested in discussing new opportunities, collaborations, and innovative projects in the AI and machine learning space.

Phone/WA: +91 9998065454

Email Me LinkedIn