A Chunked Prime Number C/Python Example - Built and Run on my own Slurm Cluster.

Event/Data Driven Architecture using Kafka, Kubernetes, OpenAPI & More.

JSON to Parquet ETL Services using Python Pandas

Ironic ML2 Driver for OpenStack Neutron, enabling bare metal provisioning with custom switch hardware profiles.

I built a complex network mesh spanning 10 geographical regions to share network and compute resources. This forms the cloud that supports my applications via Kubernetes. Not all regions have substantial servers; some are edge locations with policy-based routing to access other regions. There are 4 cloud servers, with redundancy achieved by two in NJ and two in NYC, connected via a mesh & load balanced.

Due to the current limitations of OpenStack's Magnum project, particularly its lack of support for Dual Stack, I developed my own solution to replace Magnum. Using a combination of Terraform, OpenStack, Ansible, and Cloud-Init, I built a custom bootstrap process to configure and deploy my environment and software stack independently. This cluster is now running all apps seen on this portfolio plus this site.

I configured Kafka and a dashboard in my Kubernetes environment using CI/CD automation with Ansible and the Kubernetes API. This setup tests applications requiring streaming data, like advertising tech and real-time stock feeds. Below, you can see raw data showing AAPL stock trends, sourced from a producer API and processed via a backend WebSocket for real-time updates.

Due to the complexity of Kubernetes subsystems, ingresses, and applications, developers and DevOps professionals need to monitor their environments by scraping endpoints into Prometheus time-series databases and visualizing them in dashboards. I deployed a Kubernetes-based Prometheus setup with custom tags to scrape applications with them.

I developed an advert simulator using Python, Kafka & MongoDB & more to showcase SSP and DSP marketplace ads. It streams ads based on events like page views, clicks, and purchases, serving relevant ads. Future enhancements include integrating AI (a recommendation engine) and completing the ETL chain beyond random ad sequences.

The API handles equities analysis, including volume and intraday data via real-time WebSocket. It evaluates trends and makes buy/sell calls based on stable and predicted growth. The React front end and admin portal manage equities, with predictions visualized in Grafana, offering a turnkey quant tool.

Tunninet's core product securely connects end-users to the DMZ, allowing only authenticated access while shielding the API. It also creates a decentralized marketplace for compute and AI workloads, matching microservices demand with resources, aiming to become the 'Uber of compute'.

I built my biography website because results speak louder than words. It’s a dynamic, engaging resume that offers recruiters and colleagues a comprehensive view of my achievements, providing insights beyond a traditional resume, fostering a more meaningful and lasting connection.