AVAILABLE FOR HIRE

MS DATA SCIENCE · SEATTLE U · 2026

SanjayRameshKannan

Extracting signal from noise — RAG pipelines, neural networks, and ML systems beyond notebooks.

View Projects Get In Touch

PROJECTS

ML ACCURACY

88%

RAG P@1

81.5%

LOCATION

SEA·US

ABOUT ME● LIVE

MACHINE LEARNING ◆RAG SYSTEMS ◆NEURAL NETWORKS ◆DATA SCIENCE ◆PYTHON ◆PYTORCH ◆SQL ◆QDRANT ◆SCIKIT-LEARN ◆COMPUTER VISION ◆NLP ◆DEEP LEARNING ◆MACHINE LEARNING ◆RAG SYSTEMS ◆NEURAL NETWORKS ◆DATA SCIENCE ◆PYTHON ◆PYTORCH ◆SQL ◆QDRANT ◆SCIKIT-LEARN ◆COMPUTER VISION ◆NLP ◆DEEP LEARNING ◆

// 001 — CAPABILITIES

SkillSet

DOMAINS 8

EXP 3+ yrs via coursework

Machine Learning:scikit-learnRegressionClassificationSVMsEnsemble MethodsFeature EngineeringModel Evaluation

AI & LLMs:OpenAI APIPrompt EngineeringRAGSHAPLIMENLPVector Search

Data & Statistics:PythonpandasNumPySQLRHypothesis TestingCross Validation

Visualization:TableauPower BImatplotlibseabornAltairExcelBI Dashboards

Data Engineering:AWSPySparkHadoopDockerETLGitData Modeling

Deep Learning:PyTorchOpenCVCNNArcGIS ProSpatial DataAudio ML

Cloud & APIs:AWS LambdaEventBridgeS3FastAPIVercelDocker

Analytics Infrastructure:AthenaDynamoDBS3Query PipelinesData WorkflowsCloud Integrations

// 002 — SELECTED WORK

Projects

TOTAL 6

FEATURED 5

PROJECT 01

AI / NLP

RAG SYSTEM

Football Injury RAG System

RAG pipeline combining GPS tracking, PubMed research, and player history for injury risk prediction. Engineered vector search across 50K+ points with ACWR and PlayerLoad metrics.

P@1 Score

81.5%

MRR

85.9%

Vectors

50K+

PythonQdrantPubMed APIGPS MetricsRAG

PROJECT 02

Computer Vision / Audio

DEEP LEARNING

Bird Call Classification

CNN models on Mel spectrogram inputs for audio classification. Reduced overfitting with dropout and early stopping. Full pipelines for multi-species evaluation.

Binary Acc

88%

Multi-species

57%

Method

CNN

PyTorchOpenCVMel SpectrogramsPython

PROJECT 03

Classification

HEALTH ML

Heart Attack Risk Prediction

Linear, RBF, and polynomial SVM models to predict heart attack risk. Captured nonlinear relationships among sleep, work hours, and BMI using kernel cross-validation.

Model

SVM

Eval

ROC AUC

K-Fold

scikit-learnSVMPythonSQLETL

PROJECT 04

Multi-task Learning

BEHAVIORAL ML

Predicting Youth Alcohol Use

Preprocessed 10,000 survey records from Hadoop and SQL. Solved binary, multiclass, and regression tasks simultaneously. Identified behavioral predictors via feature importance.

Records

10K

Tasks

3 types

Improved

PythonpandasHadoopSQLEDA

PROJECT 05

Full Stack / 3D

AVIATION TECH

Flight Global — Aviation Intelligence

Live 3D globe showing real-time commercial flight positions worldwide. Fleet analytics with airline stats, route density heatmaps, and live delay feeds via OpenSky Network API.

Coverage

Global

Flights

Live

Data

Real-time

Next.jsCesiumJSOpenSky APIThree.jsWebGL

// 003 — BUILT FOR FUN