Neel Mani Ratna

INDIAN INSTITUTE OF TECHNOLOGY MADRAS LinkedIn GitHub Kaggle Email Portfolio

Education
Program Institution %/CGPA Year
MTech in Computational & Experimental Mechanics Indian Institute of Technology Madras 8.6* 2025
BTech in Mechanical Engineering Techno Main Saltlake Kolkata 8.52 2021
12th CBSE Eastern Railway Inter College Jamalpur 75.8 2016
10th CBSE Saraswati Vidya Mandir Jamalpur 9.6 2014
Professional Experience
1. Working as a Freelancer to Design and Integrate Chatbot for Website
Thinkers’ media December 2024 - Ongoing
o Utilizing pre-trained GPT models (GPT-4), fine-tuning them on domain-specific datasets (depending upon the clients) to enhance
contextual understanding and optimize conversational AI capabilities.
o Integrating API-based interaction using LangChain to connect the chatbot with the language model, ensuring efficient real-time
response generation.
o Deploying and monitoring the AI chatbot using LangSmith and LangGraph, enabling scalable, observable, and structured
conversational workflows without requiring traditional front-end development.

2. Generative AI and Deep Learning for Adaptive Sheet Metal Design and Stress Prediction
Tata Steel May-July 2024
o Developed and validated a deep learning neural network model using FEM data to predict sheet metal stresses, strains, and
deformations with high accuracy, utilizing TensorFlow and supervised learning techniques to enhance computational efficiency.
o Optimized multi-pass sheet rolling operations through AI-driven predictive analytics, reducing roller stands from 8 to 6 passes,
saving $50,000 annually, and improving process efficiency.
o Applied Python, TensorFlow, GPT-based LLMs for data-driven insights, and model-based development (MBD) methodologies with
CAD to design, simulate, and enhance production control systems for scalable industrial applications.

3. AI-Powered Learning Analytics for Robotics Mentoring

Robomanthan pvt ltd Jan – June 2023
o Developed a student performance prediction model using Machine Learning Model, analyzing past assessments and
engagement metrics to identify struggling students and optimize mentorship strategies.
o Built an LLM-based chatbot using BERT and GPT models, automating responses to student queries and providing instant learning
support, reducing mentor workload by 40%.

4. Vocational Trainee
Indian Oil Corporation of India Limited Dec 2020- Jan 2021
o Worked extensively on Total Productive Maintenance (TPM) systems, focusing on optimizing the performance and reliability of
devices.
o Gained basic knowledge of the critical components and operational workflows of centrifugal compressors and gas turbines,
contributing to the development of TPM-based solutions for sustained operational excellence.
Projects
1. AI Enhanced Finite Element Modeling for stress and deformation analysis in Welded Rail Joints.
MTech Project* Phase 2, Guide: Prof K Ramesh August 2024 - Ongoing
o Utilizing Finite Element Analysis (FEA) for modeling stress and deformation in welded rail joints, employing deep learning
algorithms (e.g., Convolutional Neural Networks) to predict thermal buckling and stress-induced fractures, reducing derailment
risks.
o Integrated neural networks into the FEA framework for dynamic optimization of rail track stability parameters, combining real-
time feedback from Digital Image Correlation (DIC) with AI-powered image segmentation techniques to process and analyze
strain fields in high-resolution images.
o Working on a hybrid AI system combining supervised learning and unsupervised learning to identify key factors influencing rail
joint performance and predict potential failure points, leading to proactive maintenance strategies and reduced downtime.

2. Performance Evaluation of NLP Techniques vs. Large Language Models for Sentiment Analysis
Course Project IIT Madras
o Conducted a sentiment analysis project utilizing advanced Natural Language Processing (NLP) techniques, including TF-IDF, word
embeddings, and transformers, alongside large language models (LLMs) like GPT for text classification.
o Compared performance metrics such as F1-score, accuracy, and AUC-ROC, demonstrating that LLMs achieved almost 15% higher
accuracy over traditional NLP methods in handling nuanced linguistic patterns and contextual sentiment.
o Leveraged Python libraries such as Hugging Face Transformers, and Scikit-learn for data preprocessing (Byte Pair Encoding,
tokenization, Token Embeddings), model training, and evaluation, creating a robust pipeline for sentiment prediction and analysis.

3. Predictive and Statistical Modeling

Certified Virtual Project: PwC Switzerland, forage July-August 2024
 o Evaluated a retail banking phone marketing algorithm by predicting term deposit subscription using a dataset of 41,000 clients,
employing Ensemble methods (e.g., Random Forest, Gradient Boosting, XGBoost) and cross-validation for robust model selection
and performance evaluation with hyperparameter tuning.
o Analyzed various classification models, achieving a peak accuracy of 89.45% with XGBoost.
o Implemented feature selection techniques and dimensionality reduction using techniques like PCA and autoencoders, reducing
features from 20 to 10 with minimal loss in model accuracy (89.24%), resulting in a 28% reduction in computational cost.

4. Deep Learning for Early-Stage Cancer Detection in Medical Imaging

Course Project IIT Madras
o Fine-tuned a pre-trained deep learning model (ResNet-50) on a medical image dataset (e.g., mammograms, CT scans) for early-stage
cancer detection.
o Implemented transfer learning with a learning rate of 0.001 and utilized data augmentation techniques (e.g., rotation, flipping,
brightness adjustment) to enhance model robustness, improving generalization by 18% on a dataset of 8,000 images.
o Optimized model performance through hyperparameter tuning (e.g., batch size of 32, Adam optimizer), achieving an AUC of 0.97 and
an F1-score of 0.93, ensuring high sensitivity and specificity for early cancer detection.

5. Machine Learning Project on Real Estate Data

Personal Project March-April 2024
o Scraped and gathered 25,000 real estate data from 99acres.com, followed by data cleaning, feature engineering, and EDA.
o Applied outlier detection, removal, and missing value imputation, boosting dataset quality, and improving model performance by 14.87%.
o Engineered a machine learning model (used pipeline) for price prediction with a 93.2% accuracy rate, integrated with a
recommender system with increased user engagement.

6. Market Basket Analysis for Customer Behavior Insights

Personal Project
o Scraped and aggregated Smartprix mobile phone data from its websites using Python (BeautifulSoup, Selenium) to create a robust
dataset for analysis of customer purchasing behavior.
o Conducted exploratory data analysis (EDA) and built interactive dashboards in Power BI to visualize product associations, uncover
trends, and support targeted marketing strategies, leading to a 10% sales increase
o Automated data cleaning and preprocessing workflows, reducing manual effort by 30% and ensuring scalability for analyzing large
datasets.
Technical Skills
Programming Language: Python (OOPs, and modular coding)
Tools/Environment/Libraries: NumPy, Pandas, Scikitlearn, PyTorch, Streamlit, Seaborn, Matplotlib, SQL, Web scraping using Selenium
WebDriver and Beautiful Soup, Data Preprocessing, Feature Engineering, Feature Selection, PCA, Statistical Analysis, Machine Learning model
Training and Evaluation using pipeline, Deep Learning, LLMs, BERT and GPT, Model Pretraining (Tokenization, Byte Pair Encoding, Vector
Embeddings), RAG, Finetuning (LoRA, Quantization, QLora), LangChain, LangSmith, Ollama (llama-3.2:2b), Groq, Gen AI (DL+LLMs), MLOps
(DVC, Pipeline, MLFlow).

Courses
o Data Analytics and Visualization o Fundamental Concept of Deep Learning
o Statistics for Data Science o Introduction to Large Language Models
o Mathematical Foundation of Machine Learning
Scholastic Achievements
o Secured All India Rank of 671 in GATE 2023
Position of Responsibilities and Extra Curricular Activities
o Mentoring MTech 1st year student for the course AM5440 during the duration Jan- May 2025.
o Worked as Student Coordinator in International Conference on Experimental Mechanics (ICEM-2024) and SEM-2024.
o Mentored 220+ students in AM2540 Lab at IIT Madras, enhancing academic performance and understanding (Aug-Nov 2024).
o Gained insightful exposure of Automotive production techniques, manufacturing process including assembly and quality
control during Industrial Visit at Ashok Leyland Ennore Plant, Chennai.