Portfolio - Sahil Sainath Shirodkar

Hello, I am

Sahil Sainath Shirodkar

Data Aspirant

About Me.

Software Developer with 3 years of experience in ASP.NET Core and modern web technologies, actively transitioning into AI/ML through dedicated
self-study and hands-on projects. Completed 300+ hours of specialized AI coursework and built production-grade machine learning applications.
Seeking entry-level AI/ML role where I can apply my strong software engineering foundation to real-world data challenges.

1

Deep Learning Project

1

Machine Learning Project

1

GenAI Project

Skills.

Python

Pandas

MySQL

Angular

Pytorch

Streamlit

Machine Learning

Deep Learning

NLP

LLM

Projects.

My Experience.

Software Developer | Open Destinations Ltd, Goa | Feb 2023 - Present
Developed and maintained robust travel solutions using ASP.NET Core, Python, MySQL, and AngularJS in cross-functional teams. Optimized backend logic, reducing processing time by 99.8% (from 10 minutes to 1 second). Led critical QA and debugging efforts, minimizing downtime by 25% and reducing issue resolution time by 40%. Collaborated with QA, BAS, and support teams, boosting delivery efficiency by 20%. Recognized as key contributor for on-time delivery and supporting critical milestones.

AI Transition Initiative: Applied data analysis techniques to optimize business logic and implemented predictive features for user behavior.

Awards & Certificate.

Python: Beginner to Advanced For Data Professionals

SQL for Data Science

Math and Statistics For AI, Data Science

Master Machine Learning for Data Science & AI: Beginner to Advanced

Deep Learning: Beginner to Advanced

Natural Language Processing

Contact.

Feel free to get in touch with me. I am always open to discussing new projects, creative ideas or opportunities to be part of your visions.

https://www.linkedin.com/in/sahil-shirodkar-b520851b4/

GitHub

https://github.com/sahilshirodkar30

Deep Learning Project

Fresh Harvest Detection

Domain/Function: Agriculture / Computer Vision

Built a deep learning crop classification system using ResNet50 in PyTorch achieving 90% accuracy. Implemented image preprocessing, data augmentation, and deployed an interactive Streamlit web app for real-time harvest detection.

View Project
Engagement

View Project
on GitHUB

View Project
on Streamlit Cloud

Other Tools

PyTorch

Torchvision

Streamlit

Python

Streamlit web interface showing crop image upload and real-time prediction using the trained ResNet50 model.

Predicted output

Project Details

Fresh Harvest Detection is a deep learning–based computer vision project focused on classifying agricultural crop images to assess freshness. The system uses a ResNet50 architecture implemented in PyTorch and trained on labeled crop datasets. Extensive image preprocessing and data augmentation were applied to improve model robustness and generalization. The model was optimized using cross-entropy loss and the Adam optimizer, achieving approximately 90% accuracy.
An interactive Streamlit web application was deployed, allowing users to upload crop images and receive real-time predictions. This project demonstrates strong skills in deep learning, computer vision, and end-to-end ML deployment for real-world agricultural applications.

Machine Learning Project

Online Payment Fraud Detection

Domain/Function: FinTech / Fraud Detection

Developed a machine learning fraud detection system using Logistic Regression with SMOTE to handle class imbalance. Achieved a ROC-AUC score of 0.95 and deployed a Streamlit app for live fraud prediction.

View Project
on GitHUB

View Project
on Streamlit Cloud

Other Tools

Scikit-learn

Pandas

NumPy

Matplotlib

Seaborn

Streamlit

Streamlit dashboard displaying transaction input fields and fraud risk prediction output.

ROC curve and performance metrics highlighting the effectiveness of the fraud detection model.

Project Details

Online Payment Fraud Detection is a machine learning project designed to identify fraudulent financial transactions. The dataset exhibited high class imbalance, which was handled using SMOTE to improve minority class representation. A Logistic Regression model was trained with effective feature engineering and scaling techniques.
The model achieved a ROC-AUC score of 0.95, indicating strong predictive performance. A Streamlit-based web application was developed to allow users to input transaction details and receive real-time fraud risk predictions. This project demonstrates expertise in supervised learning, imbalance handling, evaluation metrics, and ML application deployment in a FinTech context.

GenAI Project

Medical Assistant AI (RAG-based Document Assistant)

Domain/Function: Healthcare / Retrieval Augmented Generation (RAG)

Built a healthcare-safe Medical Assistant AI using Retrieval Augmented Generation (RAG) to query medical PDFs. Implemented semantic search with HuggingFace embeddings and deployed a Streamlit-based conversational interface.

View Project
on GitHUB

View Project
on Streamlit Cloud

Other Tools

LangChain

LLaMA-3 (Groq)

HuggingFace Embeddings

ChromaDB

Streamlit

Medical Assistant AI (RAG-based Document Assistant)

Streamlit chat interface showing document upload and medical question answering workflow.

Predicted outcomes

Project Details

Medical Assistant AI is an end-to-end Retrieval Augmented Generation (RAG) system designed to answer user queries strictly based on uploaded medical PDF documents. The system uses LangChain with HuggingFace embeddings and ChromaDB for semantic search and persistent vector storage.
LLaMA-3 (Groq) was integrated as the language model, with strict context grounding to prevent hallucinations and ensure healthcare-safe responses. The Streamlit-based conversational UI supports multi-PDF uploads, document re-indexing, and chat history management. This project demonstrates strong capabilities in NLP, vector databases, GenAI systems, and responsible AI design.