DSV

DharmpratapSingh Vaghela

Data Scientist | ML Systems | Quantitative Finance | LLM Engineer

MS Data Analytics Engineering graduate at George Mason University, graduating December 2025. Currently leading ClimateGPT development as Lead Developer, processing 7.4M+ emissions records with 96.5% integration success. Specialized in LLM systems, cloud-native data solutions, and quantitative modeling.

View My Work Get in Touch Download Resume
01

Projects Award-winning ML systems and production-grade applications

FEATURED CAPSTONE
Aug 2025 — Dec 2025

ClimateGPT

Leading a 6-person team as Lead Developer to build an LLM system enhanced with climate data through Model Context Protocol (MCP) servers. Implemented MCP server with DuckDB, processing 25 years of EDGAR data (2000-2024). Built 15+ MCP tools with smart entity resolution, achieving 50% query optimization through caching. Implemented Testing Automation Framework and Baseline Knowledge Guardrails as core innovations.

7.4M+ Records Processed
100% Query Success Rate
55+ Automated Tests
Python DuckDB Docker MCP FastAPI Streamlit
🏆 1ST PLACE HACKATHON
2025 — ChallengeX Hackathon

IntegrityX

First place winner at ChallengeX hackathon. Built Next.js 14 frontend with FastAPI backend (89 endpoints across 49 modules). Integrated Walacor blockchain with 5 primitives processing documents in ~300ms. Implemented 6 fraud detection algorithms with quantum-safe cryptography (SHA3-512, AES-256) and 10-layer security stack.

1st Place Winner
~300ms Processing Time
89 API Endpoints
Python FastAPI PostgreSQL Redis Blockchain
QUANT FINANCE
2024

Bayesian Jump-Diffusion Model

Production-grade Bayesian framework implementing 9 financial models (Merton Jump-Diffusion, Stochastic Volatility, Regime-Switching, GARCH+Jumps). Built with PyMC using MCMC sampling. Delivers comprehensive risk analytics (VaR, CVaR, Sharpe ratio) at 90%, 95%, 99% confidence levels with option pricing and full Greeks calculation (Delta, Vega, Theta, Gamma).

9 Financial Models
6 Stress Scenarios
5 IC Criteria
Python PyMC Bayesian Stats Finance
MOBILE / WEB3
Nov 2022

HarmonyCustomerApp

Cross-platform Flutter marketplace for NFTs and digital artwork (iOS, Android, Web, Windows). Implemented clean architecture with Provider 6.1.1 state management and JWT authentication. Features encrypted credential storage, dual marketplace browsing, and offline capabilities with Hive 2.2.3.

4 Platforms
Clean Architecture
Dual Marketplace
Flutter Dart Provider Hive
NLP
Dec 2024 — Mar 2025

Reddit Sentiment Dashboard

Multi-model NLP dashboard analyzing Reddit posts through 5 concurrent AI models: Cardiff RoBERTa (3-class), DistilBERT (binary), NLPTown BERT (5-star), BERTweet, and emotion detection (7 categories). Optimized to 1.5-2GB RAM with Streamlit caching and shared model loading.

5 NLP Models
7 Emotion Categories
1.5GB Memory Footprint
Python Streamlit Transformers BERT
ML / FINANCE
2024

Stock Price Prediction

Production-ready NVIDIA stock forecasting with 5 ML algorithms (Random Forest, XGBoost, LightGBM, Linear Regression, LSTM). Engineered 60+ features with technical indicators (RSI, MACD, Bollinger Bands). Walk-forward validation preventing data leakage, evaluated with 15+ metrics including Sharpe ratio.

60+ Features
5 Algorithms
15+ Metrics
Python Scikit-Learn XGBoost Pandas
ML / RISK
2024

Loan Default Prediction

Credit risk pipeline achieving 0.92 ROC-AUC with 6 algorithms: Logistic Regression, Lasso, Random Forest, XGBoost, Neural Networks, SVM. Implemented SMOTE oversampling for class imbalance and PCA dimensionality reduction (95% variance retention). Refactored from notebooks to modular production architecture.

0.92 ROC-AUC
6 Algorithms
95% Variance Retained
Python TensorFlow XGBoost SMOTE Risk Analytics
EXPLAINABLE AI
May 2025

CrashAnalytics

Road accident analysis platform with CatBoost gradient boosting and model explainability. Implemented SHAP analysis and Partial Dependence Plots for safety-critical transparency. Interactive Streamlit dashboard enabling pattern discovery and risk factor identification.

SHAP Explainability
PDP Visualizations
CatBoost Model
Python CatBoost SHAP Streamlit
MOBILE / WEB3
Nov 2025

HarmonyArtist

Artist-focused Flutter platform enabling creators to sell artwork and NFTs across iOS, Android, and Web. Built with Dart (96.5% codebase) following feature-based architecture. Comprehensive testing suite with standard Flutter conventions and asset management.

96.5% Dart Codebase
3 Platforms
Feature Architecture
Flutter Dart Mobile Dev
RESEARCH
2023

Liver Disease Research

Statistical research analyzing liver enzyme correlations (ALP, ALT, AST) in hepatitis C patients. Achieved 97% accuracy with Random Forest (0.99 AUC) and 95% with Logistic Regression (0.98 AUC). Identified bilirubin as primary severity indicator through ANOVA and regression modeling in R.

97% Accuracy
0.99 AUC Score
3 Biomarkers
R ggplot2 randomForest Quarto
RECOMMENDER
2023

MovieLens Recommendation System

Movie rating prediction comparing KNN (RMSE: 1.170) and Collaborative Filtering (RMSE: 3.65). Implemented sparse matrix representation and on-demand similarity calculation for scalability. Demonstrates accuracy-scalability tradeoffs.

1.170 Best RMSE
2 Algorithms
Sparse Matrix
Python KNN Collaborative Filtering

15 Projects Total

View all repositories including additional academic projects, statistical analyses, and experiments on GitHub.

View All 15 Projects on GitHub
02

Certifications

4 Certifications

Industry-recognized credentials in data science & cloud

Google Advanced Data Analytics

Google

Aug 2024

View →

AWS Academy Data Engineering

Amazon Web Services

Jul 2024

View →

Google Data Analytics

Google

Jul 2024

View →

AWS Academy Cloud Foundations

Amazon Web Services

Apr 2024

View →
03

Experience

3 Years

Leading development teams and building production systems

Jan 2026 — Present

Quantitative Analytics Engineer Intern

Walacor

Building production-grade data integrity verification infrastructure, transforming IntegrityX—a blockchain-based document verification system created during Walacor Challenge X hackathon—into enterprise SaaS solutions. Architecting hybrid data systems combining PostgreSQL for operational queries with Walacor's blockchain technology for immutable audit trails and cryptographic proofs. Developing verified AI model infrastructure for Hugging Face marketplace, implementing SHA-256 hash attestations and tamper-proof lineage tracking. Designing RESTful APIs and SDK integrations using Python (FastAPI, SQLAlchemy) and TypeScript/Next.js for enterprise client onboarding. Creating quantitative validation frameworks for financial services compliance, including automated anomaly detection and audit trail generation.

Python FastAPI PostgreSQL TypeScript Next.js Blockchain Cryptography SQLAlchemy
Aug 2025 — Dec 2025

Model Context Protocol (MCP) Lead Developer

Team Fusion — ClimateGPT Capstone

Developed core infrastructure for ClimateGPT, an LLM system enhanced with climate data. Built the Model Context Protocol (MCP) server enabling natural-language querying of 3.4M EDGAR emissions records. Designed and implemented the Testing Automation Framework (55+ automated tests) and Baseline Knowledge Guardrails to evaluate LLM response quality.

Python MCP Server LLM Testing DuckDB Scrum
Dec 2022 — June 2023

ODOO Data Analytics Intern

Atharva Systems

Automated analysis of 500K+ ERP records, improving data processing efficiency. Optimized SQL queries reducing retrieval time by 40%. Developed Python scripts for data cleaning and preprocessing, improving data accuracy by 25%.

SQL Python ERP Systems Data Engineering
04

Education

MS + BTech

Data Analytics Engineering & Information Technology

Jan 2024 — Dec 2025

MS Data Analytics Engineering

George Mason University, Fairfax, VA

Coursework: Bayesian Inference, Machine Learning, Statistical Learning, Big Data Analytics, Database Systems, Cloud Computing, Natural Language Processing. Active in data science community and hackathons.

Machine Learning Big Data Cloud Computing
July 2018 — May 2023

Bachelor of Technology — Information Technology

Dharmsinh Desai University, India

Foundation in computer science, data structures, algorithms, and software engineering. Developed early interest in data science and machine learning through coursework and projects.

Computer Science Data Structures Software Engineering
05

Skills

30+ Technologies

Full-stack expertise across ML, cloud, and modern frameworks

Python PyTorch TensorFlow Docker AWS PostgreSQL Next.js Flutter Python PyTorch TensorFlow Docker AWS PostgreSQL Next.js Flutter
React FastAPI Scikit-learn DuckDB Git Tableau XGBoost Streamlit React FastAPI Scikit-learn DuckDB Git Tableau XGBoost Streamlit
LANGUAGES

Core Stack

  • Python
  • SQL
  • R
  • JavaScript
  • TypeScript
  • Dart
  • C++
AI / ML

ML & Data Science

  • PyTorch
  • TensorFlow
  • Scikit-learn
  • XGBoost
  • PyMC
  • LangChain
  • Hugging Face
DATA

Data Engineering

  • Pandas
  • DuckDB
  • PostgreSQL
  • Redis
  • Spark
  • Airflow
FRAMEWORKS

Web & Mobile

  • Next.js
  • React
  • FastAPI
  • Flutter
  • Streamlit
CLOUD & DEVOPS

Infrastructure

  • AWS (EC2, S3)
  • Docker
  • Git/GitHub
  • CI/CD
  • Linux
VISUALIZATION

Analytics & Viz

  • Power BI
  • Tableau
  • Streamlit
  • Matplotlib
  • Plotly
METHODS

Methodologies

  • Agile/Scrum
  • Statistical Modeling
  • Bayesian Methods
  • NLP
  • Time Series
  • Blockchain
06

Interactive Demo

Live Calculator

Production-grade Black-Scholes with real-time Greeks visualization

About: This live calculator demonstrates the Black-Scholes model for European call options. Adjust parameters to see real-time impact on option price and Greeks (Delta, Gamma, Theta, Vega, Rho). The 3D surface shows the option value landscape across strike prices and time to expiration.

Related Project: Bayesian Jump-Diffusion Model - Implements 9 financial models including Merton Jump-Diffusion and full Greeks calculation.

Gallery

Drag to explore

Hover your mouse in images and scroll to see beautiful image clock

When I'm not coding, I capture moments through photography - because sometimes images express what words cannot.