- Master of Science, Data Science | Cochin University of Science and Technology (Oct 2021 – May 2023)
- NLP Crash Course | iNeuron (June 2023 – Oct 2023)
- Masters, Machine Learning and Deep Learning | iNeuron (April 2021 -May 2022)
- Programming Data Structure and Algorithm using Python | IIT Madras (NPTEL) (Jan 2021 – Mar 2021)
- Bachelor of vocation, Software Development | St. Michael’s College University of Kerala (May 2017 – Mar 2020)
Data Analyst Intern @ To-Let Globe – Lucknow, Uttar Pradesh (June 2023 – Dec 2023)
- Uncovered insights using Python, employing Pandas and NumPy and R for efficient data manipulation
- Leveraged SQL to extract, filter, and aggregate data from databases, supporting decision-making processes
- Leveraged SQL to extract, filter, and aggregate data from databases, supporting decision-making processes
- Proficiently used Advanced Excel for data cleaning, transformation, and organizing data sets for analysis and used python and google scripts to automate sheets
- Created impactful visualizations with Matplotlib/Seaborn and synthesized findings into actionable reports
- Created insightful reports and dashboards using PowerBI and Tablue
Data Science and Python Instructor @ WISDEMY – Kochi, Kerala (part-time) (Jan 2023 – May 2023)
- Conducted engaging and interactive classroom sessions, effectively explaining data science topics and programming concepts to students
- Developed hands-on exercises, assignments, and projects to provide practical experience in applying data science techniques using Python
- Demonstrated expertise in data manipulation, data visualization, statistical analysis, and machine learning algorithms using Python libraries such as NumPy, Pandas, matplotlib, and Scikit-learn
- Implemented effective assessment methods to evaluate student’s progress and knowledge acquisition, providing constructive feedback for improvement
- Led the creation of an intelligent retrieval-based question-answering system tailored for medical queries, employing LangChain tools and methodologies.
- Curated and processed a collection of medical PDF documents using LangChain's document loaders and textsplitting techniques.
- Constructed a high-performance vector database using FAISS by extracting embeddings from the segmentedmedical texts, enhancing information retrieval efficiency.
- Incorporated Hugging Face embeddings to generate meaningful text representations, optimizing queryprocessing and retrieval within the medical domain.
- Engineered a sophisticated retrieval QA chain, leveraging language model capabilities and vector searchtechniques for contextually precise responses to medical inquiries.
- Customized the question-answering pipeline to adeptly handle diverse medical prompts and various query types,ensuring adaptability and usability for users.
- Demonstrated strong proficiency in Python and utilization of specialized libraries like langchain, Hugging Face,FAISS, and Streamlit, empowering the NLP and machine learning aspects of the project.
- Orchestrated the development of an advanced question-answering system leveraging Google Palm LLM,enabling precise responses through contextual comprehension.
- Applied robust CSV data handling techniques to extract FAQs and construct a high-performance vector databaseusing FAISS, significantly optimizing information retrieval processes.
- Integrated Hugging Face embeddings to transform text data into meaningful representations, enhancing theefficiency of query processing and retrieval.
- Spearheaded the implementation of a retrieval-based QA system proficient in generating context-awareresponses by efficiently leveraging vector search techniques and language model capabilities.
- Tailored the question-answering pipeline to adeptly handle diverse prompts and query types, augmenting itsflexibility and usability for varied user interactions.
- Demonstrated proficiency in Python, utilizing libraries like langchain, Google Palm, Hugging Face, and FAISSfor NLP and machine learning tasks. Additionally, integrated Streamlit for a user-friendly and interactive frontend experience, enhancing the accessibility of the system.
LSTM Autoencoder Based Extreme Rainfall Prediction in Highly Unbalanced Data Using Vector Reconstruction Error
- Conducted research and developed a novel approach utilizing autoencoder techniques to predict extreme rainfallin Kerala.
- Collaborated with the Advanced Centre for Atmospheric Radar Research, CUSAT, to gather and analyse highlyunbalanced rainfall data.
- Employed hyperparameter tuning to optimize the performance of the predictive model.
- Implemented regularization techniques to prevent overfitting and improve generalization.
- Devised a custom loss function incorporating mean squared error (MSE) and Kullback-Leibler (KL) divergenceterms to encourage accurate predictions and sparsity in the model's outputs.
- Utilized the ROC curve method to determine the optimal threshold for classifying rainfall events.
- Ensured a more automated and data-driven approach for classification, enhancing the reliability of thepredictions
- Developed a functional web application capable of generating captions for user-uploaded images.
- Orchestrated the backend setup using Flask, allowing seamless communication between the user interface andthe image captioning model.
- Engineered the image processing pipeline, ensuring proper handling of uploaded images and their conversion toa format compatible with the captioning model.
- Integrated the pre-trained BLIP image captioning model from the Hugging Face transformers library into theFlask application, enabling accurate generation of captions
- Optimized the model's inference process to handle multiple captions generation for each image upload,providing users with a variety of descriptive outputs for a single image.
- Contributed to the frontend design by providing necessary endpoints and functionalities for displaying uploadedimages and their corresponding generated captions.
- Developed a real-time face emotion recognition system using deep learning and transfer learning techniques.
- Utilized computer vision for image recognition and pre-processing to detect facial emotions.
- Employed transfer learning in TensorFlow using the Mobnet v2 algorithm to capture and classify facial emotions
- Utilized the AffectNet dataset, a large-scale facial expression dataset, for training and evaluation.
- Employed the Haar Cascade algorithm for face detection, enabling the identification of faces in real-time video
- Enabled the real-time expression of detected emotions on the web application's user interface using flask
- An interactive dashboard for Covid-19 analysis and prediction of India using Python and
- PowerBI Dashboard