Available for new opportunities

Uday Varikuppala

Data Scientist · ML, NLP & MLOps

5+ years building and deploying ML models across logistics, fintech, and healthcare. The kind that stay running after the demo.

Get In Touch View Work
LinkedIn ↗ GitHub ↗ Email ↗ Resume ↗
5+ Years Experience
50K+ Daily Predictions Served
3 Industries
LangChainRAGXGBoost PyTorchSageMakerSHAP SnowflakePower BI
01 Background

About
Me

5+ years building and deploying machine learning models across logistics, fintech, and healthcare. Most of that work sits in the gap between a model that looks good in a notebook and one that holds up in production: feature engineering, deployment, monitoring, and retraining when the data drifts.

Python, XGBoost, and PyTorch are day-to-day. Currently at FedEx building demand forecasting models and NLP/LLM systems, including a RAG pipeline over internal knowledge bases and a fine-tuned BERT classifier for complaint routing.

01

NLP & LLM Systems

LangChain RAG pipelines over internal knowledge bases and fine-tuned BERT models for classification and routing.

02

Forecasting & Risk Modeling

XGBoost, LightGBM, and Prophet for demand forecasting, plus survival analysis for prepayment and delinquency risk.

03

MLOps & Production

SageMaker Pipelines, Feature Store, and Evidently AI drift monitoring. Keeping models working the day after launch.

02 Capabilities

Technical
Skills

01

Generative AI & NLP

RAGLangChain Hugging Face TransformersFine-Tuned BERT NLP / LLM Applications
02

ML & Data Science

XGBoost / LightGBMTensorFlow / PyTorch Scikit-learnProphet Survival AnalysisLogistic Regression Feature Engineering
03

Languages & Data

PythonPandas / NumPy SQLR SnowflakePostgreSQL Kafkadbt
04

MLOps & Cloud

AWS SageMakerAWS Glue / Lambda / S3 MLflowAirflow DockerGitHub Actions Evidently AI
05

Explainability & Experimentation

SHAPA/B Testing Causal InferenceUplift Analysis Great Expectations
06

Visualization & BI

Power BITableauDashboarding
03 Career

Work
Experience

FedEx Corporation

Memphis, TN

Data Scientist / ML Engineer Jun 2024 – Present Current
  • Built package demand forecasting models (XGBoost, LightGBM, Prophet) on AWS SageMaker, improving volume prediction accuracy by 31% over the prior baseline and supporting hub capacity planning.
  • Developed a customer complaint classification and routing system using a fine-tuned BERT model and a LangChain RAG pipeline over internal knowledge bases, reducing escalation handling time.
  • Automated model retraining and versioning with SageMaker Pipelines and AWS Feature Store, using drift signals from Evidently AI to trigger retraining.
  • Served real-time delivery-delay risk scores via AWS Lambda and SageMaker endpoints (50K+ daily predictions at sub-60ms latency), consumed by dispatch operations for rerouting decisions.
  • Designed A/B tests and uplift analyses to evaluate customer self-service features; results informed rollout decisions.
  • Built Tableau and Power BI dashboards tracking model and delivery KPIs for operations and leadership; identified ~$18K/year in AWS cost savings.
XGBoostLightGBMProphet LangChainRAGBERTSageMakerEvidently AI

Caliber Home Loans

India (remote support for US mortgage operations)

Data Scientist, Analytics & Risk Modeling May 2021 – Jul 2023
  • Built prepayment-risk and delinquency models using XGBoost and survival analysis on behavioral and macroeconomic features to support lending and servicing decisions.
  • Developed ETL/ELT pipelines in Snowflake consolidating origination, servicing, and underwriting data for modeling and regulatory reporting.
  • Automated amortization modeling and loan-schedule generation in Python, substantially reducing manual errors and processing time.
  • Implemented Great Expectations data-quality checks for ML training and reporting datasets; optimized Snowflake queries and warehouse configuration, saving ~$50K annually.
  • Delivered Power BI dashboards with delinquency, repayment, and refinancing KPIs used by senior stakeholders.
XGBoostSurvival AnalysisSnowflake Great ExpectationsPower BI

Novartis

India

Data Scientist, Healthcare Analytics Jan 2020 – Apr 2021
  • Built patient no-show prediction models (XGBoost, logistic regression) using behavioral, demographic, and geospatial data; engineered temporal and weather features that improved model performance by ~20%.
  • Deployed a real-time scoring API (10K+ daily predictions) with SHAP-based explanations to support scheduling decisions.
  • Partnered with clinical operations teams to turn model outputs into scheduling changes and better capacity utilization.
XGBoostLogistic RegressionSHAP Geospatial MLReal-Time Inference
04 Case Studies

Featured
Projects

01

Real-Time Delivery-Delay Risk Scoring

50K+ Daily Predictions at Sub-60ms Latency
Problem

Dispatch operations needed live delay signals to make rerouting decisions, not end-of-day reports.

Solution

A real-time scoring service on AWS Lambda and SageMaker endpoints, serving delay-risk predictions consumed directly by dispatch tooling.

Impact

50K+ daily predictions at sub-60ms latency, powering live rerouting decisions.

AWS LambdaSageMakerReal-Time Inference
02

Complaint Classification & RAG Routing

BERT + LangChain RAG Pipeline
Problem

Customer complaints needed to reach the right team quickly, but manual routing was slow and inconsistent.

Solution

A fine-tuned BERT classifier paired with a LangChain RAG pipeline over internal knowledge bases to classify and route complaints automatically.

Impact

Reduced escalation handling time by connecting classification directly to the right internal knowledge and team.

BERTLangChainRAGNLP
03

Patient No-Show Prediction

~20% Model Performance Improvement
Problem

High no-show rates were hurting clinic capacity utilization, and scheduling didn't account for behavioral or environmental signals.

Solution

XGBoost and logistic regression models using behavioral, demographic, and geospatial data, with engineered temporal and weather features.

Impact

~20% improvement in model performance; deployed as a real-time scoring API (10K+ daily predictions) with SHAP-based explanations used by clinical operations.

XGBoostLogistic RegressionSHAPGeospatial ML
05 Credentials

Education &
Certifications

Master of Science in Data Science

University of Memphis

Aug 2023 – May 2025

Machine Learning · Deep Learning · NLP · Statistical Modeling · Data Engineering · MLOps

Certifications

  • Microsoft Certified: Power BI Data Analyst Associate
  • Mathematics for Machine Learning & Data Science, DeepLearning.AI
06 Let's Talk

Get In
Touch

Open to Data Scientist and ML Engineer roles, technical consulting, and research work. Particularly where forecasting and LLM systems meet production constraints.

If you're working on forecasting, NLP, or ML systems in production and want a second opinion, reach out.

Your email address is only used to reply to you. No spam, no third-party sharing.

0 / 500