Virtual and in-person experiences and curated gifting. Projects include: guestlist uploads, curated collections, shipment tracking, reservations, reusable registrations and invitations (guestlist/shared url) and RSVPs, and Admin panel views for events, guestlists, shipment tracking, reservations and event requests. Tech stack: Elixir Phoenix, Next.js/React (Typescript), Ecto, Oban, RabbitMQ, Postgres, SendGrid (mail), Twilio (SMS), Shippo (shipment tracking), and Sentry (error tracking).

Software development and consulting. Specializing in analysis, development and strategy for researchers and small tech companies. Clients include PatientPing, Splatter, Boston University, AgeTogether, Northeastern University, VocaliD, NH Lottery, MediaCrossing, Comlinkdata, GYK Antler and others. Technical highlights include the following.- Worked with neuroscience researchers at Boston University to build a framework for developing brain-computer interfaces; code, available online at https://github.com/NeuralProsthesisLab/unlock, is mostly Python and C++; technologies involved include graphics programming, machine learning and articial intelligence techniques, and signal processing.- Worked with researchers at Northeastern University on the human voicebank, https://vocalid.co/voicebank; technologies include Javascript, Python/Flask and Postgres, S3 and AWS.- Rewrote Comlinkdata's Hadoop jobs moving from Map/Reduce to Spark and introduced Ansible for conguration management and deployment automation.- Developed a data analytics platform for GYK Antler; most of the work centered around integrations with APIs including data integrations Facebook, Double Click for Advertisers, Google Analytics, Twitter, Instgram, and Pintrest as well as several other custom data feeds; technologies include SiSense, Python, Pandas, Flask, Redshift/Postgres, SQL Server.

Lead developer of software that calculates the digital metrics.

DataXu is a demand side advertising platform that applies machine learning to advertising auctions. I contributed code to the real-time bidding system, the data warehouse web service and campaign management system. Highlights include:- Designed and implemented a federated service discovery and cluster membership management system for a 500+ node computing system that spans 3 datacenters around world (Java/Scala);- Created an application for managing advertising campaigns on Facebook using the using the Facebook Ads API (Ruby on Rails and jQuery/Javascript);- Maintained the reporting warehouse web service and added several features including an administration interface, persisting and replaying requests functionality, and request failover functionality (Java).- Added asset uploading to the campaign management system. Code is based on the Sebastian Tschan file uploader (jQuery/Javascript and Ruby on Rails)

Publishing industry web development using Spring MVC, Velocity, Hibernate, Resin, Oracle, MySQL, Webistrano, Javascript and HTML. Another engineer and I did the backend and jQuery slider plugin integration for slideshows on IDG's Java-based sites (computerworld.com/slideshows, cio.com/slideshows and csoonline.com/slideshows). I also maintained the Nexus repo and the Jenkins build server.

Query RewriteAt the heart of the Akiban database is a schema-mapping module that defines an equivalence relation over the set of tables defined by the administrator. This mapping is used to transparently cluster data, and it relies on query rewriting. I developed a SQL parser and designed and implemented an algorithm for rewriting queries such that joins among elements of the same equivalence class are removed.Vertical StoreThe Akiban architecture provides high performance processing for a wide range of complex queries by combining row-oriented and column-oriented storage techniques. I developed and delivered a working prototype of a column-oriented storage engine. The design is based on Monet.

SCache: A Distributed Block-Level Flash Caching System SCache is a distributed, coherent block-level caching system. The idea is to extend the storage array onto the server by making use of server-side flash devices. I co-designed and implemented the distributed caching system in the device-mapper layer of the Linux kernel.Accelerating the Deduplication Pipeline I was responsible for evaluating the viability of using system-on-a-chip devices to accelerate computationally intensive portions of the deduplication process. I developed a benchmark-based analysis and compared benchmark performance among the following platforms: Hifn DR-255, Octeon XL Express NIC, Octeon CN5750, Intel Xeon X5355 and Intel P4. I presented the results to engineering teams and executives.Service Management FrameworkI led the development of the Service Management Framework. The framework consists of a collection of reusable interfaces and underlying software layers, which provides the infrastructure necessary to seamlessly access, control, simulate and test low-level C-based abstractions from a Python interface.Clustering ArchitectLead designer of a Linux-based clustering and replication service. At the core of the design is a notion of separation between system clustering of storage processing elements and pair based replication of I/O Objects.

Software Defined RadioI was part of a four person scrum team that over the course of 12 sprints delivered a C++-based iDEN Base Radio for Sprint-Nextel. Additionally, I worked on a National Science Foundation grant, Programmable Radio Platforms for Highly Dynamic Networks.

UNIX ClusteringI participated in the development cycle of three releases of Tru64 UNIX, and I was the maintainer of 3 core cluster kernel modules, including the Internode Communications Subsystem, the Distributed Request Dispatcher, and the Reflective Memory driver. Projects included product enhancements, prototypes, white papers and presentations to executives, engineering teams and customers. Other daily duties included core-dump triage, bug fixes and analysis, and high priority customer patches.