Master of Science — Management Information Systems

Aug 2024 – May 2026

University of Illinois Chicago (UIC)

Relevant coursework: Data Mining for Business, Machine Learning & Statistical Methods, Business Data Visualization, Deep Learning & Modern Applications, ML Lifecycle & Deployment.

Bachelor of Technology — Electronics & Communication Engineering

Aug 2018 – Jul 2022

Dayananda Sagar University, Bengaluru, India

AWS Certified Cloud Practitioner

Amazon Web Services

Research — Sustainability Research & Innovation Congress

Spanning 80+ countries

AI platform for benchmarking municipal climate investments adopted by local governments.

INFORMS Hackathon 2026

Built a two-agent AI system solo in 24 hours.

IEEE Published Author

"Heart Rate Variability and ML for Stress Analysis" — DOI: 10.1109/I4C57141.2022.10057650

Agentic AI & LLM

LangChain · LangGraph · DSPy · CrewAI · RAG pipelines · ChromaDB · hybrid retrieval · prompt engineering · MCP · Anthropic Claude API

LLM Evaluation

RAGAS · MLflow · LLM-as-judge · drift detection · golden test sets · regression testing

Machine Learning

Python · scikit-learn · XGBoost · LightGBM · CatBoost · PyTorch · SHAP · DoubleML · DiD · PSM · Bayesian modeling (PyMC) · A/B testing · anomaly detection

Data Engineering

SQL · dbt · Snowflake · BigQuery · PostgreSQL · Fivetran · Airflow · Dagster · Kafka · PySpark · Databricks

Backend & Infrastructure

FastAPI · Docker · Kubernetes · GitHub Actions · CI/CD · AWS (Certified) · Azure · GCP

Visualization

Tableau · Power BI · Looker · Streamlit · Plotly · Recharts

Graduate Research Assistant — Agentic AI

May 2025 – May 2026

University of Illinois Chicago  ·  CLEAN Initiative (Prof. Selvaprabu Nadarajah)

• Built a production agentic AI system scoring Illinois municipalities against a climate sustainability framework — adopted by local governments to benchmark climate investments.
• Designed an end-to-end ML data ingestion pipeline collecting and structuring heterogeneous data (sustainability plans, emissions inventories, census data) across 100+ municipalities.
• Built a prompt-engineered LLM document parsing system achieving 95%+ extraction accuracy across heterogeneous municipal documents.
• Architected a 4-stage multi-LLM orchestration pipeline: DSPy ReAct agent (5 tools), hybrid BM25 + dense retrieval, and OutputCritic self-verification layer.
• Built a RAGAS evaluation suite gating every release on 12 automated checkpoints across 7 prototype generations; raised retrieval precision 42%.
• Achieved a GradientBoosting scoring model with LOO MAE 5.01 across 27 municipalities. Tracked 100+ experiments with MLflow.
• Presented findings to the Metropolitan Mayors Caucus and urban planning groups. Research accepted at the Sustainability Research and Innovation Congress (80+ countries).

Data Analyst

Aug 2024 – May 2026

University of Illinois Chicago  ·  Online Programs Team

• Analyzed behavioral and engagement data across 15,000+ students and 500+ online courses on UIC's LMS/Blackboard platform.
• Built Power BI and Tableau dashboards translating engagement data into decisions for academic stakeholders.
• Maintained data quality and governance standards across all downstream reporting.

Data Scientist

Aug 2022 – Aug 2024

PlanSource  ·  Benefits Administration SaaS  ·  Bengaluru, India

• Built a churn prediction system using XGBoost and LightGBM (0.84 ROC-AUC across 15K segments) from 40+ behavioral features drawn from COBRA, eligibility, enrollment, and remittance data; quantified $2.1M in annual revenue at risk and directly changed the product roadmap.
• Built ETL pipelines using Python and Fivetran connecting Salesforce, NetSuite, Google Ads, and LinkedIn Ads into Snowflake, with dbt models at 98% test coverage.
• Performed monthly reconciliation audits across COBRA, eligibility, and EDI 834 files, identifying mismatched records and mapping errors before client-facing delivery.
• Integrated LLM-powered automation via the Anthropic API with structured output validation.
• Applied causal inference (DoubleML, synthetic control, propensity score matching) to isolate true effects from confounded behavioral signals.
• Reduced manual processing by 15–20 hours per week through automation; created SOPs supporting onboarding of 37 additional team members during a workload surge.