ExperienceImpactSkills ProjectsEducationResearch CertificationsContact
HS Research Tech 2 - University of New Mexico, Albuquerque

SaagarikaSampath Kumar

Data Scientist and AI Engineer working at the intersection of biomedical research and production ML. I fine-tune single-cell foundation models, build spatial transcriptomics pipelines, and turn complex signals into measurable, interpretable outcomes across omics, clinical AI, NLP, and LLMs.

0+
Years Experience
0+
Projects Shipped
3.93/4
Graduate GPA
0
Publications
single_cell_embedding.umap
Spatial Omics Deep Learning NLP / LLMs Applied ML
Scroll to explore
FIG. 01

Professional Journey

Jun 2026
- Present
HS Research Tech 2
IM Epidemiology & Biostatistics, University of New Mexico
Albuquerque, NM
  • Developing AI/ML pipelines in Dr. Vadigepalli's lab: fine-tuned scSimilarity foundation models and deployed CellViT and Path2Space on HPC clusters for cell segmentation and spatial mapping across multi-omics single-cell datasets.
  • Integrated Visium and Xenium spatial transcriptomics with H&E image analysis to co-register morphological and gene-expression signals, enabling spatially resolved biomarker discovery.
Foundation ModelsSpatial TranscriptomicsSingle-Cell AIHPCBiomarker Discovery
Feb 2026
- May 2026
Visiting Research Assistant
UNM MIRAI Lab, University of New Mexico
Albuquerque, NM
  • Developed an explainable ML framework for ordinal falls-risk prediction using markerless motion-capture data, extracting clinically interpretable biomechanical features from sit-to-stand assessments in older adults.
  • Designed an end-to-end AI/ML pipeline spanning skeletal motion preprocessing, feature engineering, subject-independent validation, model explainability (SHAP), and fairness analysis for transparent risk stratification.
Explainable MLMotion CaptureSHAPClinical AIFairness Analysis
Sep 2025
- Dec 2025
Data Science Intern, Analytics & Automation
Media Sales Plus
Buffalo, NY
  • Built and deployed ARIMA and LSTM forecasting models across 3+ product lines using Python and SQL, delivering production-grade revenue analytics with 95% confidence intervals.
  • Developed end-to-end data pipelines (scikit-learn, SQL) reducing manual effort by 40%, and created Tableau and Streamlit dashboards that accelerated leadership decision cycles by 30%.
40% Less Manual Work95% CI Forecasts30% Faster Decisions
Jul 2023
- Jul 2024
Data Analyst
Saint-Gobain Research India
Chennai, India
  • Cleaned and transformed 2M+ weekly records across 5+ source systems using PySpark and SQL, building robust pipelines and conducting EDA that surfaced ~$40K in operational inefficiencies.
  • Deployed LSTM models (PyTorch) on 50+ IoT sensor streams and applied GANs to synthesize 10K+ structured product design variants, partnering with cross-functional teams on high-impact analytical problems.
~$40K Savings Surfaced10K+ Designs Generated50+ IoT Sensors
Dec 2022
- Jul 2023
Audit Data Science Intern
Ernst & Young LLP
Chennai, India
  • Designed and deployed a supervised ML fraud detection system (Python, scikit-learn, SQL) processing 25K+ payslips at 92% precision across 100+ client engagements.
  • Engineered NLP and OCR pipelines extracting 50K+ unstructured financial records and automated 15+ Alteryx workflows, eliminating 120+ manual hours across the analytics function.
92% Fraud Precision15+ Pipelines120+ Hours Saved
FIG. 02

Measurable Impact

Fraud Detection
0%

Precision on a supervised ML fraud detection system processing 25K+ payslips at EY.

Ernst & Young - 2022/23
Cost Optimization
~$0K

In operational inefficiencies surfaced via PySpark/SQL EDA across 5+ Saint-Gobain systems.

Saint-Gobain Research - 2023/24
Efficiency
0%

Reduction in manual reporting effort via automated ML pipelines at Media Sales Plus.

Media Sales Plus - 2025
Data Extraction
0K+

Unstructured financial records extracted via NLP and OCR pipelines to automate risk insights.

Ernst & Young - 2022/23
Model Accuracy
0%

Classification accuracy on a Vision Transformer fine-tuned on the Cats vs Dogs dataset.

University at Buffalo - 2025
Decision Speed
0%

Faster leadership decision cycles from interactive Streamlit and Tableau dashboards.

Media Sales Plus - 2025
FIG. 03

Technical Expertise

I work across the full stack of biomedical and applied AI - from single-cell data and spatial omics through model deployment and LLM-powered insight layers.

Currently focused on
Single-Cell Foundation ModelsSpatial TranscriptomicsMulti-Omics AIClinical AI Systems
Bioinformatics & Spatial Omics05
scRNA-SeqSpatial Transcriptomics (Visium)XeniumscSimilarityPath2Space
Languages & Data Engineering08
PythonRSQL (PostgreSQL)SQL (MySQL)PySparkPandasNumPyETL Pipelines
ML & Statistical Modeling05
Supervised LearningUnsupervised LearningClassificationRegressionTime Series Forecasting
Deep Learning & Computer Vision07
PyTorchTensorFlowCNNRNNLSTMGANImage Segmentation
NLP & Generative AI06
LLMsRAGFine-TuningPrompt EngineeringHugging FaceLangChain
BI & Visualization06
TableauPower BIStreamlitPlotlyAlteryxExcel (Advanced)
FIG. 04

Selected Projects

01Spatial Omics
Spatial Transcriptomics Integration Pipeline

Co-registers Visium and Xenium spatial transcriptomics with H&E image analysis, deploying scSimilarity, CellViT, and Path2Space on HPC clusters to link tissue morphology with gene expression for spatially resolved biomarker discovery.

Foundation ModelsVisium / XeniumCellViTHPC
02Biomedical AI
Explainable Falls-Risk Scoring from Motion Capture

End-to-end pipeline converting markerless sit-to-stand motion data into an interpretable ordinal falls-risk score for older adults, with biomechanical feature engineering, subject-independent validation, SHAP explainability, and a fairness audit.

Pythonscikit-learnSHAPMotion Capture
03LLM / RAG
Clinical Literature Copilot - RAG Assistant

Retrieval-augmented assistant for querying biomedical papers in natural language, built with Hugging Face embeddings, a vector store for semantic search, and LangChain orchestration returning grounded, citation-linked answers.

RAGLangChainHugging FaceEmbeddings
04NLP / GenAI
Gen AI in Education - UB Data Challenge

NLP pipeline for prompt generation and summarization using Python, TensorFlow, and Hugging Face. Increased dataset diversity by 30% and annotation precision to 90.2%, earning an honorable mention from the jury.

PythonTensorFlowHugging FaceNLP
05Medical NLP
AI Radiology Report Summarizer

Retrieval-augmented GRU model summarizing chest X-ray reports via TF-IDF chunk retrieval on the OpenI dataset. Deployed as an interactive Streamlit app for clinicians with image-linked outputs.

RNN/GRUTF-IDFStreamlitRAG
06Time Series
Production Revenue Forecasting - ARIMA & LSTM

Forecasting system across multiple product lines combining ARIMA and LSTM for 12-month revenue projections with 95% confidence intervals, packaged into Streamlit and Tableau dashboards that cut decision cycles by 30%.

ARIMALSTMStreamlitTableau
FIG. 05

Academic Background

Graduate Degree - 2025
Master of Science
Data Science & Applications
University at Buffalo, SUNY
Aug 2024 - Dec 2025
GPA3.93 / 4.0
Deep Learning - Advanced NLP - Cybersecurity - Database Management - AI & Statistical Data Mining
Integrated Master's - 2023
Master of Science
Decision & Computer Sciences
Coimbatore Institute of Technology
Jul 2018 - May 2023 - Minor in Statistics & Data Science
GPA3.85 / 4.0
Computer Vision - Machine Learning - Statistics - Python & R Programming - Financial Analysis
FIG. 06

Research Papers

Research
Securing Pharmaceutical Supply Chains with Ethereum Blockchain
CCIS, Springer - 2025
View
Research
Stock Market Analysis via MapReduce & PySpark
IRJMETS
View
Research
Automobile Safety System
International Journal of Innovative Research in Engineering
View
FIG. 07

Licenses & Certifications

Anthropic
AI Fluency: Framework & Foundations
Anthropic - 2026
Verify
Anthropic
Claude 101
Anthropic - 2026
Verify
Anthropic
Claude Code 101
Anthropic - 2026
Verify
AWS
Building a Generative AI-Ready Organization
Amazon Web Services - 2025
Verify
AWS
Planning a Generative AI Project
Amazon Web Services - 2025
Verify
Google
Fundamentals of Digital Marketing
Google
View
Google
Explore ML Beginner Track
Google
View
Microsoft
Introduction to Azure Fundamentals
Microsoft
View
Cert
Impact of AI and ML
St. Joseph's Institute of Technology, OMR - 2020
View
Udemy
The Complete Ethical Hacking Course
Udemy - 2020
Verify
Cert
Introduction to Digital Marketing
North Storm Academy (formerly Socialopedia)
View

Let's build
something meaningful.

Seeking full-time roles in Data Science, ML Engineering, and Biomedical AI. Drawn to teams building intelligent systems that bridge research and real-world impact. Open to relocation across the United States.

Send a Message