Data Scientist & ML Engineer • Available for Impact

Building Intelligent Systems & Agentic AI Architectures

Hi, I'm Pathik Rupwate. With 4+ years of hands-on experience and an MS from the University of Chicago, I architect and deploy end-to-end predictive models, Generative AI agentic workflows (MCP, RAG, LangChain), and computer vision pipelines that turn complex datasets into measurable business growth.

Test Interactive Labs View Case Studies Download Resume
Pathik Rupwate
UChicago MS • ML
UChicago
MS in Analytics / Applied ML
5 Labs
Live In-Browser GPU Inference
Agentic AI
MCP, LangChain & Gemini RAG
Prod ML
Computer Vision & Quant Systems
// Core Competencies

Technical Expertise & Tooling

Comprehensive toolset spanning production machine learning, Generative AI agent architectures, and enterprise analytics.

Generative AI & Agents

Large Language Models (LLMs) RAG Architecture Model Context Protocol (MCP) LangChain Gemini AI ElevenLabs API Agentic Workflows API Integration

Machine Learning & Modeling

Time-Series Forecasting Anomaly Detection Natural Language Processing (NLP) Computer Vision (VGG-16) Causal Inference Classification & Regression Clustering (K-Means, DBSCAN) GANs & SMOTE

Programming & Libraries

Python PyTorch TensorFlow & Keras scikit-learn Pandas & NumPy R (ggplot2, caret) SQL C++ Java

Infrastructure & MLOps

Docker Flask APIs GCP (Google Cloud) AWS Linux GitHub Copilot MongoDB

Analytics & BI

A/B Testing Tableau Dashboards Power BI Multivariate Testing Heatmaps & Time-Series KPI Alignment

Domain & Leadership

FinTech Risk Modeling Government MCP Automation Executive Presentation Cross-Functional Agile Student Gov President ($250k) Public Sector / GovTech Ad Strategy Simulations
// Live In-Browser AI Engine

Interactive Machine Learning & Agent Labs

Test live, browser-native implementations of production predictive models, computer vision neural networks, and autonomous multi-tool agent workflows.

Credit Risk Modeling • Isolation Forest • Explainable AI (SHAP Waterfall) • German Credit Benchmark

Applicant Credit & Risk Parameters

Evaluate how unsupervised Isolation Forests and ensemble models detect high-risk loan defaults under severe class imbalance. Adjust credit metrics or select preset profiles to observe real-time scoring.

Quick Applicant Presets:
Credit Score (FICO) 680
Debt-to-Income Ratio (DTI %) 35%
Loan-to-Value Ratio (LTV %) 75%
Monthly Transaction Volatility Index 28
Past 2-Year Delinquent Occurrences 0
Anomaly Score
18%
Standard Risk (Approved)

Multi-Dimensional Risk Radar & Local Feature Attribution

Explainable AI (XAI) breakdown decomposing the net anomaly score into individual Shapley feature contributions relative to the baseline expected risk ($\mathbb{E}[f(x)] = 22\%$).

SHAP Local Feature Attribution (φ) Base: 22.0% → Net: 18.0%
Quantitative Macro Modeling • S&P 500 Sector Alpha • Extra Trees Ensemble (78.85% Accuracy)

Macroeconomic Regime Simulation & Sector Alpha

Explore how macroeconomic inflation shocks and Federal Reserve interest rate shifts drive sector rotation across all 11 S&P 500 sectors, augmented with Tabular GANs to model tail-risk regimes.

1-Click Macro Scenarios:
Macro CPI Inflation Shock (+/- %) +2.5%
Fed Interest Rate Trajectory (bps) +50 bps
GAN Synthetic Samples
25,000 pts
Model Robustness Lift
+20.0%

Sector Outperformance Probability (%)

Predictive outperformance probabilities across all 11 S&P 500 sectors generated by Extra Trees classifiers under the active macroeconomic regime.

Top Alpha Pick
Energy (XLE)
Valuation Drag
Tech (XLK)
Model Accuracy
78.85%
Computer Vision & Deep Learning • MIT-CSAIL Benchmark • 468 3D Mesh + VGG-16 Architecture

Real-Time 3D Facial Mesh & Neural Inference

High-precision facial geometry extraction via a 468-point 3D landmark mesh, feeding 22 calibrated anthropometric ratios into a fine-tuned deep convolutional neural network (VGG-16 architecture).

Test 1-Click Benchmark Faces:
Slim Face
Slim / Lean
BMI ~20.5
Normal Face
Normal
BMI ~25.2
Fuller Face
Fuller Structure
BMI ~33.8
Camera Positioning Guidance
For best results and optimal landmark alignment, face directly at the camera with your head upright, eyes open, and good front-facing lighting.
Active: Normal / Athletic (Sample B)

Live Neural Network Telemetry

Real-time geometric ratios and continuous regression output running directly in-browser with WebGL GPU acceleration.

Live Predicted BMI
25.2
Normal / Athletic Range (18.5-25.5)
Tracking Engine
MediaPipe 468-Mesh
Inference Backend
TFJS WebGL (GPU)
Tracked Face Area
210 x 235 px
Pipeline Latency
14.2 ms
Model Context Protocol (FastMCP) • Sub-50ms LLM Reasoning

Autonomous Multi-Tool Agent Studio

Interactive demonstration of an autonomous enterprise agent utilizing the Model Context Protocol (FastMCP) for multi-tool planning and execution. Select an operational mission below to trigger the reasoning and dispatch loop.

Public Revenue Intelligence Agent 21 FastMCP Tools
Audits municipal utility billing telemetry across districts, flags high-value delinquent accounts, and generates statutory recovery notices.
AppSec Vulnerability Triage Agent DAST + Pydantic
Executes dynamic application security testing (DAST) on payment APIs, verifies SQL injection vectors via LLM, and synthesizes validated remediation code.
Intelligent Document Processing (IDP) Agent Vision CNN + OCR
Classifies legal deeds with fine-tuned MobileNetV2, extracts land parcel metrics via OCR, and validates title records against cadastral registries.
Tool Architecture
get_water_top_defaulters_ward
Execution Latency
36.4 ms (Groq)

Live ReAct Trace & Structured Synthesis

Follow the agent's internal thought stream, real-time FastMCP JSON protocol dispatch, observation ingestion, and final structured output.

Unsupervised Customer Intelligence • K-Means & PCA • Silhouette & Elbow Optimization • NLP Topic Modeling

Unsupervised Behavioral Clustering

Interactive demonstration of multi-dimensional customer segmentation combining PCA dimensionality reduction with K-Means clustering to identify high-value cohorts and optimize targeting.

Select Domain Dataset:
Silhouette Quality
0.68 (Strong Separation)
Optimal Clusters
K = 4 (Elbow Validated)
Inertia Reduction
-68.8% WSS Reduction

Behavioral Persona & Topic Modeling (LDA)

Select any cluster in the scatter plot to inspect its financial profile, latent Dirichlet allocation (LDA) review topics, and targeted campaign recommendations.

Cluster 1: High-Net-Worth Savers +34% Campaign Lift

High balance preservation, low credit risk, high interest in wealth management and municipal bonds.

NLP Review Topics (LDA Extraction):
yield stability wealth advisory fixed deposit
Avg. Balance
$142,500
Conversion Rate
18.4%
// Portfolio & Case Studies

Featured High-Impact Projects

Key engineering, modeling, and generative AI initiatives delivering quantifiable business and operational impact.

GenAI & Agents ABM Knowledgeware

Municipal AI Agent (Mainet Upgrade)

Flagship product feature transformation

Spearheaded the development of a municipal AI agent utilizing the Model Context Protocol (MCP), Retrieval-Augmented Generation (RAG), and complex Data APIs. Integrated this agent as a major feature upgrade to flagship Mainet software, significantly accelerating critical data accessibility for government officials.

MCP RAG Data APIs Python LLMs
Automation & AI ABM Knowledgeware

Automated Localization Pipeline

Saved 80+ hrs production time / cycle

Directed the end-to-end architecture of an automated, multi-lingual tutorial video generation pipeline using Python and the ElevenLabs API, completely eliminating manual voiceover work and streamlining product rollouts across languages.

Python ElevenLabs API Multilingual NLP Video Ops
Data Ops & Recovery ABM Knowledgeware

Large-Scale Data Restructuring

80% recovered from 400K+ files

Engineered a sophisticated data parsing and recovery solution to salvage 400,000+ unstructured files recovered from an AWS incident. Developed custom pattern-matching algorithms to restore original directory structure and metadata, averting critical data loss.

Python AWS Incident Recovery Pattern Matching Data Pipelines
AI Security ABM Knowledgeware

AI Vulnerability Tester & Auto-Patch

Autonomous remediation snippets

Designed and implemented an AI-driven vulnerability testing suite by integrating OWASP ZAP with Gemini AI, enabling automated discovery of security flaws alongside intelligent, autonomous generation of remediation code patches.

Gemini AI OWASP ZAP Python Auto-Remediation
FinTech & Forecasting Aiolux

Financial Sector Forecasting with GANs

80% Predictive Accuracy (200-day horizon)

Predicted S&P 500 sector performance across 50, 100, and 200-day horizons. Generated synthetic economic data via Tabular GANs and SMOTE to resolve severe class imbalance (+20% robustness). Identified macroeconomic drivers (CPI) via causal inference.

GANs SMOTE Extra Trees Causal Inference
Banking & NLP Kearny Bank

Customer Segmentation & Loan Default Detection

+30% ROI & +25% Risk Detection

Built customer persona clustering (K-means, DBSCAN, PCA) combined with NLP topic modeling (LDA/TF-IDF) on user feedback to drive +30% ad campaign ROI. Deployed Isolation Forests and One-Class SVMs to boost rare loan default forecasting by 25%.

Isolation Forests One-Class SVM Clustering NLP (LDA) Tableau
Computer Vision University of Chicago

Real-Time BMI Prediction Pipeline

Real-time Live Video Stream Inference

Engineered an end-to-end real-time BMI prediction pipeline utilizing a VGG-16 CNN deep learning architecture. Served the model in production via a Flask API processing live video input feeds with OpenCV optimization.

VGG-16 CNN TensorFlow OpenCV Flask API
GenAI & Chatbots Live Gemini & In-Browser Agent

AI-Powered Portfolio Chatbot

Dual Engine: In-Browser & Gemini Live

Interactive AI chatbot embedded directly in this website! Features dual-engine architecture: an instant in-browser semantic knowledge engine with optional direct Google Gemini Live LLM integration.

Gemini API Client RAG JavaScript LLM Reasoning
// Career History

Professional Experience

Track record of building data-driven systems across enterprise technology, banking, and academic research.

Data Scientist / Machine Learning Engineer

ABM Knowledgeware
Mar 2026 – Present • Mumbai, India
  • Municipal AI Agent (Mainet Upgrade): Spearheaded development of an AI agent utilizing Model Context Protocol (MCP), RAG, and complex Data APIs for the flagship Mainet software, dramatically improving data accessibility for government officials.
  • Automated Localization Pipeline: Directed end-to-end architecture of an automated, multi-lingual tutorial video generation pipeline using Python and ElevenLabs API, saving over 80 hours of production time per release cycle.
  • Large-Scale Data Restructuring: Salvaged 400,000+ unstructured files recovered from an AWS incident; developed pattern-matching algorithms to accurately restore original directory structure and metadata for 80% of the dataset.
  • AI Vulnerability Tester: Integrated OWASP ZAP with Gemini AI to automate discovery of security flaws and autonomously generate remediation code snippets.
  • Odisha Government Agent (Ongoing): Leading the initiative to build and deploy a specialized, highly scalable AI agent to modernize regional administrative tasks for millions of residents.

Data Scientist

Kearny Bank
Oct 2023 – Jan 2026 • Austin, TX
  • Improved marketing campaign ROI by 30% through advanced customer segmentation (K-means, hierarchical clustering, DBSCAN, PCA, Naive Bayes, Random Forest), optimizing ad targeting.
  • Increased forecast accuracy by 25% for loan default predictions using Isolation Forests and One-Class SVM architectures on imbalanced datasets.
  • Performed sophisticated NLP on unstructured user feedback using TF-IDF and LDA to extract actionable insights and model user sentiment.
  • Conducted rigorous A/B and multivariate testing with Product Managers to align data initiatives with business KPIs.
  • Designed dynamic executive dashboards in Tableau and Power BI utilizing heatmaps, time-series charts, and decision trees.

Data Science Researcher

University of Chicago
Jan 2023 – Jun 2023 • Chicago, IL
  • Built and deployed a real-time BMI prediction pipeline utilizing a VGG-16 CNN architecture served via a Flask API processing live video input.
  • Optimized facial recognition models for complex video analysis using TensorFlow and OpenCV, reducing latency and enhancing inference performance.

Data Scientist

Aiolux
Jan 2022 – Dec 2022 • Chicago, IL
  • Predicted S&P 500 sector performance across 50, 100, and 200 days using GANs, SMOTE, and Extremely Randomized Trees, achieving an average accuracy of 80%.
  • Generated synthetic economic data to resolve severe class imbalance, increasing overall model robustness by 20%.
  • Identified macroeconomic drivers (CPI) using causal inference techniques and applied findings to predictive ad strategy simulations.

Database Engineer

Beloit College Powerhouse
Aug 2020 – May 2021 • Beloit, WI
  • Designed, architected, and implemented a robust database and front-end website to track 5,000+ inventory items using SQL, Python, and HTML.

Software Engineer

Open Energy Dashboard
Jan 2020 – Dec 2020 • Beloit, WI
  • Built interactive diagrams and implemented multilingual architecture for an open-source energy dashboard utilizing Docker, Ubuntu, and React.
// Academic Background

Education & Leadership

Strong foundation in applied data science, computer science, and quantitative economics.

Graduated Jun 2023

Master of Science, Applied Data Science

University of Chicago • Chicago, IL
Focused on advanced machine learning, deep learning architectures, computer vision, and scalable statistical inference.
Graduated May 2021

Bachelor of Science, CS & Quantitative Economics

Beloit College • Beloit, WI
Leadership: President, Beloit Student Government — Managed and allocated a $250,000 annual funding budget for student initiatives and campus organizations.
// Let's Connect

Ready to build high-impact AI solutions?

Whether you are seeking a Data Scientist / ML Engineer for agentic AI architectures, predictive models, or end-to-end data pipelines, let's talk.

pathikrds@gmail.com +91 9321194480 linkedin.com/in/pathik-rupwate pathikrupwate.com github.com/pathik10
Mumbai, India
Send an Email Download PDF Resume