Projects

Portfolio

Production Systems & Applied Research

Featured Projects

Production Systems & Applied Research

Production · ML
Vendor Invoice Anomaly Detection System

Production anomaly detection pipeline at University of Rochester combining rule-based heuristics with Isolation Forest to flag fraudulent or erroneous vendor invoices automatically — eliminating manual review overhead.

Isolation Forest Python SQLAlchemy Alembic MLflow Docker
$50K/year cost savings at University of Rochester
  • Hybrid rule-based + Isolation Forest pipeline with automated alerting
  • SQLAlchemy ORM + Alembic for schema evolution and data lineage
  • MLflow experiment tracking for model versioning and performance monitoring
  • Dockerized for reproducible deployment on university infrastructure
Production · Data Engineering
Research Analytics Platform

Comprehensive analytics platform at University of Rochester processing 800GB+ of institutional research data to support data-driven hiring decisions and impact assessments.

BigQuery dbt Dash Tableau Python LLMs
+20% research media coverage · 800GB+ data processed
  • Processed 800GB+ institutional data for strategic decision-making
  • Integrated LLMs to automate 30% of data preparation tasks
  • Deployed interactive Dash dashboards to Posit Connect
  • Built peer benchmarking views for institutional strategy
Production · Risk Analytics
Risk Strategy Simulation Tool

Automated risk analysis tool at T-Mobile that revolutionized credit risk management, boosting team productivity by 40% and reducing investigation time by 75%.

Python Pandas Snowflake Databricks SAS→Python
+40% team productivity · 75% reduction in investigation time
  • Automated risk strategy simulations end-to-end
  • RCA-inclusive alerts for policy deviation detection
  • Migrated legacy SAS pipeline to Python for reproducibility & MLOps
Production · Fintech ML
XGBoost Credit Risk Models

End-to-end credit risk ML platform at RedCarpetUp (YC S15) serving 500k+ users — from feature engineering on 5TB data to production deployment with Docker and Flask.

XGBoost A/B Testing Docker Flask PostgreSQL Airflow
+40% MoM credit disbursal growth · ₹5.2M revenue identified
  • Led cross-functional teams to 40% MoM credit disbursal growth
  • XGBoost models with rigorous A/B testing across 4,000+ customers
  • Docker + Flask deployment with monitoring and alerting
  • EDA on 5TB data to surface ₹5.2M revenue opportunities
Research · NLP
Expansive Insights: Developer ChatGPT Analysis

Capstone research project analyzing 107,000 developer interactions with ChatGPT to understand GenAI's impact on developer productivity and identify distinct usage patterns.

Python NLP Statistical Analysis Data Mining Pandas
  • Processed 107,000 developer interactions with advanced filtering & deduplication
  • Identified 16 single-turn and 8 multi-turn interaction types
  • Applied statistical methods to analyze productivity patterns
Research · Deep Learning
Mini-GPT: Transformer Implementation

Built a Transformer-based language model from scratch using PyTorch, implementing attention mechanisms and training on the Tiny Shakespeare dataset.

PyTorch Transformers Deep Learning NLP
  • Implemented multi-head attention and positional encoding from first principles
  • Trained on Shakespeare corpus for character-level text generation
  • Achieved coherent character-level language modeling
Production · Platform
Enterprise CRM System

Built a comprehensive CRM at RedCarpetUp serving 500k+ users with KYC image processing, audit trail database schemas, and card management for 100k MAU.

80% of KYC verification automated · 100K monthly active users
Flask PostgreSQL Python REST APIs Image Processing
  • Scalable CRM for 500,000+ users with full data management
  • KYC image processing automating 80% of verification workflow
  • Database schemas for audit trails and regulatory compliance
  • Led team of 4–5 engineers through sprint cycles and delivery

Impact & Results

Quantifiable Outcomes from Production Systems

$50K
annual cost savings
Vendor invoice anomaly detection at University of Rochester
40%
MoM growth
Credit disbursal growth led at RedCarpetUp via ML-driven strategy
75%
investigation time reduction
Via RCA-inclusive alert system at T-Mobile risk management
80%
KYC automated
Image processing pipeline for document verification at RedCarpetUp
500K+
users served
Production ML platform and CRM system at RedCarpetUp
30%
efficiency gain
LLM-assisted data preparation automation in research pipelines

Research & Publications

Peer-Reviewed Work

2017
Peer Reviewed
IEEE SmartTechCon
Bangalore, India
Peer-to-peer Knowledge Sharing Platform for Farmers with Auto-Recommendation Feature

Developed an intelligent recommendation system for agricultural knowledge sharing, enabling farmers to access relevant information and best practices through a peer-to-peer platform with automated suggestion capabilities.

Recommendation Systems Machine Learning Agricultural Tech Python
DOI: 10.1109/SmartTechCon.2017.8358498

Let's Build Something

Open to ML Engineering & Data Science Roles

Start a Conversation View Full Experience