Portfolio - Pravat Nagarkoti

Hello, I am

Pravat Nagarkoti

AI/ML Enthusiast

About Me.

I'm a dedicated Data Scientist with a passion for AI, Machine Learning (ML), Deep Learning (DL), and Data Science. My interests include exploring and implementing Transformers, Computer Vision (CV), Natural Language Processing (NLP), and Large Language Models (LLMs) to build intelligent, scalable solutions. With expertise in Python, SQL, Power BI, and Tableau, I specialize in transforming complex datasets into actionable insights, predictive models, and interactive dashboards. I also develop agentic AI applications and automated workflows using LangChain, LangGraph, and n8n, enabling intelligent orchestration and enterprise-ready AI systems.

Core Expertise:
• Machine Learning Models & Algorithms
• Data Analytics & Statistical Modeling
• Interactive Dashboards (Power BI, Streamlit, Excel)
• Data Visualization, Insights & Storytelling
• Agentic AI & Orchestration (LangChain · LangGraph · n8n)

Skills.

Python

SQL

Excel

Power BI

Tableau

Machine Learning

LangChain

LangGraph

Projects.

My Experience.

Data Science Intern – Cognifyz Technologies

Dec 2024 – Jan 2025

Preprocessed restaurant data (missing values, inconsistencies, data types); ran EDA with heatmaps, bar plots, and interactive maps to show customer preferences and usage patterns.
Engineered features, applied label encoding, built ML models to predict outcomes, evaluated using RMSE and R² for performance measurement.
Collaborated with a team to develop ML workflows, distributing tasks for preprocessing, feature engineering, and model evaluation.

AI Fellow – Fusemachines

Dec 2024 – Jan 2025

Completed modules on AI/ML, data wrangling (Pandas/SQL), visualization, linear & non-linear models, ensembles, NLP, CV, time series, and LLMs.
Built and evaluated ML/DL models with performance metrics; applied CI/CD (GitHub Actions) for automated training and tuning.
Developed agentic AI apps using LangChain/LangGraph with prompt engineering, RAG, and workflow automation.

Awards & Certificate.

Python: Beginner to Advanced For Data Professionals

SQL for Data Science

Math and Statistics For AI, Data Science

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.

Download CV

[email protected]

Call Me

+977 9818782660

https://www.linkedin.com/in/pravat-nagarkoti

GitHub

https://github.com/pravatn01

Visualingo – Images to Captions and Voice with AI

Domain/Function: Computer Vision & Multimodal AI

Developed an AI-powered web app that converts images into captions and audio using a dual-model pipeline: DenseNet201+LSTM (TensorFlow) and BLIP Transformer (Hugging Face). Features include Gradio UI, gTTS voice synthesis, and dynamic model comparison.

View Project
on GitHUB

View Project
on Web

Other Tools

Deep Learning Project

Visualingo – Images to Captions and Voice with AI

Gradio web interface showcasing dual-model caption generation and audio playback from uploaded images.

Notebook code showing caption output from DenseNet201 + LSTM model on a sample image.

Project Details

Visualingo is an advanced image captioning web app that translates visual content into text and speech using AI. Built using TensorFlow and Hugging Face, the app leverages two parallel models—CNN+LSTM and BLIP Transformer—for generating descriptive captions.

It features a smooth Gradio interface that allows users to upload images, view generated captions, and listen to the spoken version via Google Text-to-Speech (gTTS). The project emphasizes multimodal interaction and showcases model outputs, improving accessibility.

Key Takeaways:
✅ Integrated TensorFlow (DenseNet201+LSTM) and PyTorch (BLIP) in a unified pipeline.
✅ Implemented Gradio for a clean, responsive web UI.
✅ Used gTTS to convert generated captions into human-like speech.
✅ Displayed output examples from both traditional and transformer-based models for comparison.
✅ Optimized inference pipelines with preprocessing and batch captioning.
✅ Containerized setup with environment.yml for reproducibility.

FinTrack – Expense Tracking Web App

Domain/Function: Personal Finance & Expense Management

Developed a robust expense tracking web app to streamline financial management, providing seamless CRUD operations, insightful visualizations, and automated testing for reliability. This tool enhances financial awareness and budgeting with data-driven insights.

View Project
on GitHUB

Other Tools

Python Project

Interface for inserting, updating, and viewing expense data stored in the database, built with Streamlit and FastAPI.

A line chart displaying monthly expense trends, accompanied by a detailed table for in-depth expense analysis.

Project Details

FinTrack is a modern web application built with FastAPI and Streamlit, it offers an intuitive interface for users to record, manage, and analyze expenses. The app provides real-time insights through dynamic visualizations and ensures data accuracy with automated testing.

Leveraging FastAPI for a high-performance backend, Streamlit for an interactive frontend, and Plotly for insightful analytics, this project showcases expertise in full-stack development, API design, data visualization, and software testing.

Key Learnings & Technical Highlights:
✅ Mastered ETL workflows for structured expense tracking.
✅ Built a scalable backend with FastAPI for efficient CRUD operations.
✅ Developed an interactive frontend using Streamlit for real-time expense analysis.
✅ Implemented dynamic charts (Pie, Line) with Plotly for financial insights.
✅ Integrated automated Pytest-based testing to ensure data reliability.
✅ Implemented logging for backend API calls.

Churn Analysis and Prediction Dashboard

Domain/Function: Telecom Churn Analysis & Retention Strategy

Developed an interactive Power BI dashboard to analyze customer churn patterns and predict at-risk new joiners. The project provides telecom companies with data-driven insights to enhance customer retention strategies.

View Project
Engagement

View Project
on GitHUB

View Project
on Web

Other Tools

Power BI Project

Displays customer churn patterns, highlighting high-risk states, plans, and demographics for retention strategies.

Predicts at-risk new joiners, focusing on high-risk plans, payment methods, and demographics to guide retention efforts.

Project Details

This Power BI project analyzes customer churn and predicts at-risk new joiners for telecom companies, enabling targeted retention strategies.

📊 Key Insights
✅ Churn Rate: 27.9% (1,073 of 3,844 customers)
✅ High-Risk Plans: Fiber Optic highest, Cable lowest
✅ Contract Impact: Higher churn in month-to-month plans
✅ Demographics: Higher churn among females, 45+ age group, and tenure >24 months
✅ Prediction: 229 new joiners at risk

🚀 Retention Strategies
🎯 Loyalty incentives for DSL & Fiber Optic users
🎯 Discounts for price-sensitive customers
🎯 Promote long-term contracts
🎯 Targeted campaigns for Texas & New York

⚙ Workflow
🛠 SQL Server for ETL, Power BI for transformation
🧠 Random Forest model in Jupyter Notebook
📈 Interactive dashboards in Power BI

This project highlights BI, predictive analytics, and retention strategy expertise.

Excel - Driven YOY Sales Growth Tracker

Domain/Function: E-commerce Analytics & Performance Optimization

Built an interactive Excel dashboard to analyze year-over-year (YOY) sales performance (2021 vs. 2022) for Hakoot, a US-based e-commerce platform. The tool enables data-driven decision-making by visualizing key sales trends, regional performance, and category contributions.

View Project
Engagement

View Project
on GitHUB

Other Tools

Excel Project

A dynamic, interactive dashboard showcasing key sales performance metrics, providing insights into year-over-year trends and regional performance.

Utilizes pivot tables, KPIs, and various charts to present in-depth, data-driven analysis for strategic decision-making and performance comparison.

Project Details

This project analyzes year-over-year sales trends for Hakoot, a U.S.-based e-commerce platform. Built in Excel, the dashboard provides actionable insights by comparing 2021 vs. 2022 key metrics.

📊 Key Insights
✅ Sales Growth: 2022 revenue rose 21.33%, profit surged 38.73%, and quantity sold increased 25.76%.
✅ Top-Selling Categories: Technology (36%), Office Supplies (33%), Furniture (31%).
✅ Regional Leaders: California, New York, Texas drove the highest sales.
✅ Seasonal Trends: Peaks in November & December, with steady growth over 2021.

🚀 Strategic Actions
🎯 Expand in top states with targeted promotions.
🎯 Optimize inventory for high-profit categories.
🎯 Boost sales in low-performing months.

📈 Dashboard Features
✅ Automated Data Cleaning
✅ Dynamic KPIs & Slicers
✅ Advanced Visualizations

This project highlights expertise in BI, trend forecasting, and data-driven strategy using Excel.

Deep Dive - PostgreSQL Insights

Domain/Function: Music Store & Sales Data Analysis

Conducted comprehensive SQL-based data analysis on a Music Store and Sales dataset using PostgreSQL. Applied advanced SQL techniques such as aggregations, joins, CTEs, and subqueries to extract meaningful insights into sales performance, customer behavior, and inventory trends.

View Project
on GitHUB

Other Tools

SQL Project

Visual representation of the music database schema, including tables for artists, albums, tracks, genres, invoices, customers, and playlists.

Overview of the sales database schema, featuring tables for customers, delivery persons, pin codes, orders, and products.

Project Details

This SQL project showcases the power of PostgreSQL for data-driven decision-making through two distinct analyses:

1️⃣ Music Store Analysis:
Imported a full database to analyze sales patterns, customer preferences, and inventory management. Used GROUP BY, aggregations, and joins to derive insights, and created an Entity-Relationship Diagram (ERD) to understand database structure.

2️⃣ Sales Analysis:
Processed CSV-based sales data to assess trends and product performance. Leveraged CTEs and subqueries for detailed analysis, and developed an ERD to optimize query performance and visualize data relationships.

Key Learnings & Technical Highlights:
✅ Designed and optimized SQL queries for structured data analysis.
✅ Applied aggregation, joins, and CTEs to extract business insights.
✅ Utilized PostgreSQL & pgAdmin for database management and ERD generation.
✅ Conducted trend analysis to enhance decision-making in sales and inventory.