I worked on many parts of Chrome OS but the primary focus was user login, user session initialization and startup. This is a huge flow which spans from from secure boot, TPM, platform daemons through the Chrome C++ code all the way up into the HTML and JS.I built many parts of the flow. Added performance metrics, Perfetto traces. Did performance analysisand improved the on-device code speed on all levels from the system IO to the main C++ code andgraphics performance.I designed and implemented architecture that allowed us to switch the user login flow implementationfrom the old statically precompiled code to modern component-based Polymer Material design runtime architecture improving system speed and ability to adapt to device performance.I designed a Family (child) accounts management on-device flow for Chrome OS. It spans from the Google Cloud Enterprise device management to the on-device code. Child account support is now one of the core features of Chrome OS.A lot of work was done in writing device performance tests in Go language and optimizing Chrome C++ code. Major issues were improvements in IO bottleneck on user session start, CPU congestion when loading many resources in parallel and improving pixel rendering.

Chrome OS

I did two main courses:- Telecommunication technologies- Perl programming languageMy experience with first building a telecom company and then developing management and provisioning systems for the large IT infrastructure resulted in my deep passion for networking and infrastructure. I was thrilled when I got a chance to teach Telecommunication Technologies at Moscow State University so I could design a curriculum to introduce students to different network technologies through the full OSI stack starting from synchronous and asynchronous networks, frame switching, Ethernet, IP, L3 switching, MPLS, internet routing etc.Another course on Perl programming language was a result of my experience integrating this language into large systems when appropriate. It was fun and I hope students learned enough to use it in their professional life.I learned a lot from colleagues and students and I hope they l found my courses useful.

"Target" advertising project (https://www.facebook.com/myTargetGlobal/):- user action analysis and prediction.- optimal advertisement selection.- advertising logs collection and real-time analysis.It was a fun time serving tens of thousands personally targeted ads per node per second. I built a real-time user targeting statistics and planning daemon. I learned how to use bloom filters, summing bloom filters, how to serve ads auctions and how to collect real time statistics and present it to the customers in real time to tune their ads.I created the first public versions of libsvm with Nvidia CUDA support. (https://github.com/alemate/libsvn-cuda)

I built a full search cluster management system that deployed binaries and data, planned load, did runtimemanagement and monitoring. We had tens of thousands of physical servers in distributed geographical locations that were wired into a single large cluster that was load-balancing the worldwide search engine that greatly improved cost-performance ratio and resource utilization.My system was responsible for planning configuration updates, distributing the binaries and data, solving the split-network configuration issues, etc. We could also automatically revert the global setup to the last known good configuration with respect to (potentially) partial previous rollout. And this all was automatically scheduled in steps to minimize the user impact.

I led the team that designed and implemented scalable Telecom / ISP / Datacenter managementsoftware. Our solutions allowed Mobile / ISP companies to provide IP-based services to their users. Itsupported all kinds of services starting from basic ISP traffic management and billing solutions, email,web services and up to data streaming services like a video streaming on a Pay-Per-View andPay-Per-Minute basis.This provided an invaluable chance to look at how large telecom companies work. How to build scalable management solutions for distributed service networks. And how to avoid building custom solutions for each company making scalable and flexible universal management software that can be supplied to different customers.We built a full ISP/Hosting services management solution with individual user control, provisioning and billing to fully automate Telecom companies.

Our company built one of the largest regional LANs in Moscow at that time. As a CTO I wasresponsible for full design, installation and support of the regional network. This included doing legalpaperwork, system design, evaluation of available network technologies and a lot of research.Designing network topologies, cable routing, network design, service management and billing.In basically two years we built a distributed city network with a 10+ kilometer radius covering a few city districts (a few city counties in US terms). Developed, built and fully certified custom IP routers, implemented full services provisioning and set up billing.Set up the company structure, schedule, policies, departments, shifts, tasks flow, etc...