FULLSTACK &AI

High-throughput Kafka ingestion + exactly-once semantics

Streams stock ticks from MySQL into Kafka partitions and persists atomic Parquet batches to HDFS with checkpoint-based crash recovery.

Fault-tolerant ingestion with tunable consistency

Processes NOAA weather streams in a distributed Cassandra cluster with configurable read/write consistency and automatic failover behavior.

Spark SQL + ML pipelines + LLM query interface

Analyzes large-scale CodeContests data with Spark RDD/DataFrame/SQL APIs, optimized bucketing/caching, and natural-language query translation.

MySQL to HDFS pipeline with gRPC failover logic

Builds partition-aware analytics workflows with replication-aware recovery, intelligent caching, and production-style gRPC APIs.

Multi-threaded processing + GIL benchmark analysis

A high-performance concurrent word-count pipeline with configurable thread pools and benchmark suites comparing GIL/no-GIL scaling plus CSV, Parquet, and Arrow format performance.

Posterior probability classification for English vs Spanish

Implements Bayes theorem over letter-frequency distributions to classify shredded text as English or Spanish, with a clear inference pipeline from character counts to language posterior scores.

Unsupervised country segmentation + hierarchical cluster analysis

Clusters country-level indicators into interpretable groups using unsupervised learning and dendrogram visualizations, making it easier to compare development patterns and macroeconomic similarity.

Normalized data pipeline + train/eval loops + top-3 prediction output

Builds FashionMNIST loaders with normalization, trains baseline and deeper feedforward PyTorch models using SGD + cross-entropy, evaluates test accuracy, and prints top-3 class probabilities for inference.

Regression benchmarking with feature engineering and residual analysis

Compares Linear, Ridge, Lasso, Random Forest, and XGBoost regressors on real-estate data after correlation-guided feature selection, log transforms, scaling, and diagnostic residual plotting.

Production retrieval + generation workflow

End-to-end RAG system for RFI/RFP automation with quality-focused retrieval and practical delivery through a usable product interface.

Automation startup website and product surface

Built a polished, conversion-friendly web experience with production-ready architecture and performance-focused frontend implementation.

Investor-startup matching and analytics platform

Combines scraping, filtering, and visual analytics to help users discover startups through data-backed ranking and exploration.

Inventory-aware consumer planning app

Transforms available ingredients into practical meal planning workflows with a clean user experience and rapid frontend interactions.

PCA-based computer vision pipeline

Implements dimensionality reduction and eigenface techniques for image compression and face-analysis experimentation.

Graph-powered navigation engine

Developed a Java-based map and routing system for UW-Madison, applying graph-theory algorithms for pathfinding at campus scale.