Work

Event-driven middleware that syncs commerce platforms with several ERPs through a pluggable adapter pattern: one sync engine, a thin adapter per system. RabbitMQ publishing with dead-letter queue handling, Redis caching, a REST API layer, and a PHPUnit suite running in GitHub Actions.

Carried product data from PIM out to the systems downstream of it on RabbitMQ and Kafka, covering inventory catalogs in more than 20 countries. Tightened the sync and validation logic across B2B and B2C product data, cutting data-integrity errors by 70%.

Tuned a 1M+ SKU catalog with Varnish, Redis, Elasticsearch, and CDN caching plus an indexing strategy, improving page load for 2,000+ daily users. Also handled production support: debugging incidents across the platform, the message queues, and the downstream integrations, with New Relic for monitoring.

Custom middleware connecting 5–6 disparate ERP systems, NetSuite, SAP, and ERPNext among them; cleansed the data at the source as it arrived from the frontend and kept downstream systems synchronized in real time. Deployed on AWS ECS and GCP GKE.

Low-latency ingestion and streaming analytics over Apache Kafka and Faktory, with dashboards for live social-media data.

A panoptic-segmentation model for robot navigation in dynamic environments, integrating perception with path planning.

Continuous American Sign Language recognition from IMU rings and virtual IMU data, built on wearable sensing and sequence models.

Generates, stores, and uses rainbow tables across a distributed architecture coordinated by a central server. Written in Rust.

A custom RL environment with a Q-learning agent and an evaluation harness for measuring policy quality.

A convolutional neural network for handwritten-digit classification built from scratch in Python and NumPy.

Proposed to create a small system which can ingest misinformation in social-media posts by combining NLP models for text with CNNs for images and quite accurately prodict if it was fake/misleading

Signal-strength, tower-switching, and network-stability study around Binghamton using Android data and Power BI.

Packet-capture analysis of VPN traffic to assess encryption and identify plaintext/PII leaks.