Available for new opportunities

Uday Varikuppala

Senior Data Scientist · GenAI & Causal ML

6+ years building production-grade LLM, RAG, and Causal ML systems across Fintech, Healthcare, and Enterprise AI.

Get In Touch View Work
LinkedIn ↗ GitHub ↗ Email ↗
6+ Years Experience
$4M+ Revenue Impact
3 Industries
LangChainLangGraphRAG GraphRAGPyTorchPinecone XGBoostSageMaker
01 Background

About
Me

Senior Data Scientist with 6+ years building and evaluating foundation and generative AI systems — LLMs, RAG, knowledge graphs — to improve product quality and user experience. Expert in designing model evaluation frameworks and benchmarks, including LLM-as-a-judge workflows, statistical testing, and failure analysis to surface edge cases and improve robustness.

Proficient in Python, PyTorch, TensorFlow, and production ML tooling. Strong technical-reporting and stakeholder communication skills to translate evaluation results into product decisions. Proven track record in Causal ML — applying CausalForest, SHAP, and uplift modeling to uncover bias, latent drivers, and actionable improvements across deployed models.

6+ Yrs Experience
$4M+ Revenue Impact
3 Industries
01

GenAI & LLM Systems

RAG, GraphRAG, multi-agent workflows, LLM-as-a-judge evaluation, and prompt engineering at scale

02

Causal ML & Explainability

CausalForest, SHAP, uplift modeling, and A/B experiment design to uncover latent drivers and reduce bias

03

MLOps & Production

SageMaker Pipelines, CI/CD, model versioning, and observability for reliable production ML systems

02 Capabilities

Technical
Skills

01

Generative AI & RAG

RAG / GraphRAGMulti-Agent LangChain / HaystackLLMs (OpenAI, Anthropic) Agentic AIPrompt Engineering Knowledge GraphsHugging Face Semantic SearchVector Embeddings
02

ML & Data Science

XGBoost / LightGBMTensorFlow / PyTorch Scikit-learnDeep Learning NLP / NLUReinforcement Learning Bayesian NetworksAnomaly Detection OptunaFeature Engineering
03

Languages & Data

PythonPandas / NumPy SQL / PostgreSQLSnowflake R / ScalaApache Spark KafkaAirflow
04

MLOps & Cloud

AWS SageMakerDocker / K8s MLflowCI/CD (GitHub Actions) SageMaker Feature StoreVertex AI GCP / Azure MLTerraform
05

Causal & Responsible AI

SHAP / LIMECausalForest Uplift ModelingA/B Testing ML Bias AnalysisExplainable AI Latent Driver Analysis
06

Visualization & BI

Power BITableauStreamlit Plotly / MatplotlibDashboarding PineconeRedisFastAPI
03 Career

Work
Experience

Quicc

Memphis, TN

Senior Data Scientist / ML Engineering Aug 2023 – Present
  • Built cross-domain evaluation framework stitching behavioral, engagement, and semantic signals; drove a statistically significant 25% lift in conversation completion rate
  • Architected RAG, multi-agent workflows, and LangChain knowledge-graph pipelines for semantic QA and conversational AI; cut R&D cycle time for new LLM features by 30%
  • Reduced data access latency by 40% via embedding large text corpora and generative SQL workflows on AWS Glue and PostgreSQL
  • Engineered SageMaker Pipelines with AWS Feature Store for automated model versioning and retraining, reducing model downtime by 60%
  • Applied SHAP and CausalForest for failure analysis, uncovering bias, underspecification, and latent drivers in deployed GenAI models
  • Built Tableau and Streamlit dashboards translating complex GenAI metrics into executive narratives used in two funding rounds
  • Mentored 2 junior data scientists and 1 intern on RAG architectures, causal reasoning, and ML lifecycle best practices
LangChainRAGMulti-Agent SageMakerSHAPCausalForestStreamlit

Caliber Home Loans

Hyderabad, India

Data Scientist — Retail Financial Decision Science & ML Engineering May 2021 – July 2023
  • Built cross-domain ML frameworks for mortgage risk, prepayment prediction, and underwriting eligibility using causal reasoning and statistical modeling
  • Deployed production ML models via Kafka-driven microservices and CI/CD workflows, cutting end-to-end pipeline latency by over 50%
  • Applied SHAP-based explainability and causal reasoning to underwriting workflows, improving classification confidence with bias-aware model governance
  • Migrated legacy batch ETL into scalable Spark- and Kafka-driven pipelines, reducing data processing time by over 50%
  • Built Power BI dashboards delivering live predictive KPIs for senior risk and finance leaders
  • Implemented Great Expectations data quality checks and confidence-interval monitoring to detect drift and regressions
XGBoostSHAPKafka Apache SparkPower BIGreat Expectations

Novartis

Hyderabad, India

Senior Data Scientist — Healthcare Decision Science & Clinical AI Jan 2020 – April 2021
  • Built cross-domain learning framework for clinical scheduling combining behavioral, geospatial, and demographic signals, recovering an estimated $1M+ in revenue through AI-driven operational decisions
  • Applied causal reasoning and engineered temporal, weather, and geospatial features to improve model evaluation metrics by ~20% over baseline
  • Deployed production-grade real-time inference API with FastAPI serving 10,000+ daily predictions at sub-50ms latency
  • Used SHAP-based XAI to surface model behavior insights for clinical teams, translating outputs into actionable operational changes
FastAPISHAPCausal ML Geospatial MLReal-time Inference
04 Case Studies

Featured
Projects

01

AI-Powered Patient No-Show Prevention

$1.2M Annual Revenue Recovery
Problem

Manual scheduling led to high no-show rates. Rule-based systems couldn't incorporate dynamic signals like weather or geospatial proximity.

Solution

Healthcare NLP & ML pipeline combining appointment history, weather data, and geospatial patterns via a real-time FastAPI inference service on AWS.

Impact

10,000+ daily predictions at sub-50ms latency. $1.2M annual revenue recovery.

XGBoostFastAPISHAPGeospatial MLAWS
02

Customer Lifetime Value Prediction

$2.7M ROI Opportunity Surfaced
Problem

Inability to predict customer churn and LTV in fintech mortgage operations resulted in missed retention opportunities and poor marketing spend.

Solution

Fintech ML engine using K-Means clustering, survival analysis, and gradient boosting to model churn probability and LTV segments with Optuna tuning.

Impact

$2.7M ROI opportunity surfaced. Reduced manual processing errors by 70%.

K-MeansSurvival AnalysisLightGBMOptuna
03

Intelligent User Journey Analytics

$430K Monthly Revenue Loss Stopped
Problem

Undetected platform bugs causing silent revenue loss across millions of e-commerce sessions. Standard monitoring missed browser-specific anomalies.

Solution

NLP pipeline processing millions of user interaction logs with anomaly detection and causal inference to surface hidden behavioral patterns.

Impact

Identified Safari bug causing $430K/month revenue loss — enabling immediate fix.

Anomaly DetectionCausal InferenceKafkaNLP
05 Credentials

Education &
Certifications

🎓

Master of Science in Data Science

University of Memphis

Top 1% GPA · Currently Enrolled

Machine Learning · Deep Learning · GenAI · LLMs · RAG · Causal ML · MLOps · Cloud AI

📜

Certifications

  • Microsoft Certified: Power BI Data Analyst Associate
  • Advanced Google Analytics — Google
  • Mathematics for ML & Data Science — Coursera / DeepLearning.AI
  • Programming Essentials in Python — Cisco Networking Academy
06 Let's Talk

Get In
Touch

Open to senior IC roles, technical consulting, and research collaborations at the intersection of LLMs and production systems.

Whether you're building a new AI product, evaluating a deployed model, or need a senior technical voice — I'd love to hear about it.

uday.v3669@gmail.com

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

0 / 500