Hi, I am Tejan, Data scientist with a strong background in statistical analysis and mathematics. I have extensive experience in solving real-world problems with the help of AI and Machine learning.
I am problem-solving oriented with a keen ability to learn quickly and adapt to diverse settings.
I can help you with every aspect of the data science workflow, including business understanding, data exploration, modeling, validation, deployment and monitoring.
My business intuition and communication skills allows me to make a quick impact and deliver high-quality results.
Find me here:
Supervised and unsupervised machine learning algorithms, performance evaluation and validation, packaging and deployment
Univariate and bivariate analyzes, hypothesis tests, treatment of missing values and outliers
Textual data processing, topic modelling, text classification, sentiment analysis
Implementation of deep learning algorithms using neural network architectures
Portfolio
Deploying a Flask web application on Azure Web App, that allows residents of my apartment to shop from the grocery store located within the apartment complex using Flask, HTML, CSS, JavaScript, and MongoDB.
This is a PDF merging tool that uses the Kivy framework.
Deployed the Flask
web application
(AWS Beanstalk + CodePipeline) to scrap reviews from Flipkart using Beautiful Soup.
A Selenium based approach for scraping images from Google.
PyWebIO-based recommender system of sci-fi multimedia or entertainment, which includes movies, series, anime for users of the program who are indecisive or just lazy to pick one up.
A Django web application to give categorized (small/quick, medium, or large) YouTube's best recipes using the YoutubeSearch API based on the ingredients provided by the user. This was later deployed on PythonAnywhere.
Building a GUI application using PyQt5 that can help the farmers to make better decisions on the crops to be harvested based on several surveys and parameters stored in the SQLite database.
Built, deployed, and ran a containerized Linear Regression model (web app) on AWS App Runner using Docker, AWS (EC2 & ECR) for diamond price prediction.
MLflow to log all the parameters and metrics into the database.
A content based movie recommender system (using NLTK library) by applying bag of words technique. Subsequently deploying on Heroku and served through Streamlit.
To build an app using FastAPI for classifying trucks as having APS-related component failure or not by utilizing Kafka for data streaming, Azure services for deployment, GitHub Actions for automation, S3 bucket for data storage, finally performing predictive modeling with XGBoost Classifier.
Built an app to predict the risk of thyroid from lab data collected on various patients from automated Clustering and by utilizing XGBoost Classifier to achieve great accuracy.
Create a Random Forest model that can help identify whether newly registered complaints are problematic or not using
PySpark for efficient data processing. Monitoring and visualization through Grafana, InfluxDB, and Prometheus, with Airflow orchestrating seamless workflows and CircleCI for deplyment.
Trained a VGG16 CNN model to classify whether chicken is coccidiosis infected or healthy through fecal images. DVC was employed to facilitate tracking and automation of pipeline. Azure services such as container registries and app services coupled with GitHub Actions for seamless CI/CD deployment.
Designed an embeddings-based image search engine employing technologies such as PyTorch and ResNet-18 for generating embeddings, Annoy Algorithm for efficient neighbor search, MongoDB for data storage, S3 for model and image with public ACL, Paperspace for GPU-based training and model registry, FastAPI for the prediction endpoint.
Leveraged cutting-edge technologies to ensure secure and efficient access control. The system employs NodeMCU, MTCNN, FaceNet, MongoDB, Docker, Azure, and Terraform to create a robust authentication mechanism.
Created a text summarization model using Hugging Face Transformers and PyTorch as a framework for building and training the neural network architecture. The FastAPI endpoints was finally deployed on Heroku.
Built a model that aims to identify and delineate individual cells within images. Roboflow aided in annotating images into a suitable format. YOLOv8 provide bounding box coordinates around each cell. Azure Container Registry stores Docker container images, while Azure App Services enables deploying and managing web applications.
Follow online to find out when this launches.
Landing Page Software