Eugene Miloslavsky Email & Phone Number

Mountain View, Ca, Us

I'm contributing to DNAnexus' high YoY ARR growth with low attrition rate by helping the users of the DNAnexus Titan platform and UK Biobank Research Analysis Platform to∙ Use Workflow Description Language (WDL), Common Workflow Language (CWL) and DNAnexus interfaces to define bioinformatics analysis∙ Analyze large genomic datasets using tens of thousands of CPUs in a cost-efficient, reliable and reproducible manner∙ Utilize curated, performant and secure bioinformatics tools provided by DNAnexus∙ Perform interactive exploratory and data science work by using DNAnexus-provided CLI, RStudio and JupyterLab applications∙ Monitor and optimize cloud resource utilization of large-scale parallelized genomic analysis workflows∙ Improve security stance when using DNAnexus platform and applicationsMy efforts go towards∙ Driving product requirements based on overall company direction and feedback from prospects, customers, internal R&D and sales organizations∙ Prioritizing projects based on tactical and strategic objectives and contributing to the product roadmap for the company∙ Working closely with engineering to coordinate cross-team dependencies and trade-off product specs, implementation constraints and timely delivery in user-facing and cloud-facing areas.∙ Defining product delivery milestones and updating C-level executives on project progress∙ Generating monthly internal training content for the new product functionality∙ Keeping track of competitive and adjacent products on the market

∙ Swivl's cloud video platform helps K12 teachers improve. I jumpstarted company-wide InfoSec process based on ISO 27001 helping Swivl close business.∙ DotScience provides cloud-based machine learning environment supporting transparent versioning for data and code, experiment tracking and collaboration. As a product manager, I conducted market research and helped simplify and re-focus the product based on my ML experience.

Criteo is a leader in ML-based, ROI-optimized advertisements, serving a trillion retargeting display ads in 2017 for 8 billion products.Improved ML product classification into a hierarchical taxonomy of 2000 categories based on product description and image.• Constructed 3 million product image training set covering 10 countries, used it to train and optimize a Keras/Tensorflow ResNet50 image classification convolutional neural network, then combined image classifier with existing text-based classifier for a 6% reduction in error rate.• Improved calibration of the existing text-based classifier.• Implemented different hierarchical text classifiers in Keras/Tensorflow that were on par with optimized liblinear-based implementations.• Built a text classifier based on FastText embeddings trained on 100M unlabeled product descriptions for minor accuracy improvements.Improved feature selection and model parameters for generalized linear model for bid prediction in Google AdWords shopping campaigns optimizer product.Finished a 3 month machine learning bootcamp consisting of lectures, programming assignments and a project where I evaluated and improved a delayed conversions model.Implemented, optimized Spark and Hadoop data pipelines.Architected granular metrics library adopted by multiple teams.Tools: Python, Tensorflow, Keras, liblinear, Spark, Scala, Java, Hadoop, graphite, grafana, scikit-learn, jira, gerrit, Kafka, Vertica, MS SQL.

Mountain View, Ca, Us

DNAnexus is a HIPAA-compliant, cloud-based genomics platform enabling data storage, large-scale, flexible analysis and collaboration on a 20,000 cpu/Petabyte scale. I designed and implemented flexible, dynamic AWS instance lifecycle management (handling 10,000 instances / day), zero-downtime deployment process, and numerous system features helping DNAnexus onboard its first major academic, pharma and clinical test companies. Mentored and lead a cloud devops team. Contributed to system monitoring, troubleshooting, CI, AWS reservations budgeting, application logic enhancements, rate limiting, security enhancements, development and release process definition in a fast-paced, agile, collaborative environment. Extended DNAnexus platform into multiple AWS regions and into AWS-China.Tools: Node.js, python, bash, streamline.js, Ubuntu, upstart, AWS EC2 (spot, reserved, EBS, VPC, Direct Connect), S3, Chef, zookeeper, mongo, redis, nginx, fluentd, sumologic, Jenkins, zabbix, dynect, , openVPN, iptable, git, jira

San Jose, California, Us

Content Repository (CR) product allowed for horizontal scaling based on a cluster of NetApp filers. I contributed to the design, implementation and stabilization of SnapMirror block-based data protection for CR resulting in the first beta release of CR SnapMirror. Tools: agile, C++, python, Unix, WAFL, SnapMirror

Wilsonville, Or, Us

Mentor Graphics is a leading provider of design automation tools for IC and electronics industries. While at Mentor, I’ve assumed increasing levels of responsibility, architecting or contributing to three successful products that enable advanced chip production at Intel, IBM, AMD and Panasonic among others.• Co-architected OPCVerify product (lithography simulator and verification checks), implemented majority of initial functionality producing the "record level of success for a new product introduction, booking 10 million in the first 10 months." (9/19/2006 Mentor Graphics official news release) • Enabled customer base growth and enhanced product functionality by collecting requirements and presenting proposals during face-to-face technical and management review meetings with key customers in North America and Asia.• Led, trained and mentored a growing engineering team. Translated customer requirements into detailed technical specifications, managed tight schedules and execution plans for the team. Insured code quality and stability with tools and policy for thorough testing and documentation, reducing the number of customer-critical issues 4X over 5 years.• Improved scalability to allow efficient use of 1000-CPU clusters (up from 80 CPUs) with optional hardware-based acceleration.• Improved the speed (~2x) and accuracy (~2x) of the lithographic simulator by utilizing signal processing, architectural, algorithmic and low-level optimization techniques. • Designed and implemented efficient classification and prioritization tool for reducing error review time for billions of errors. • Authored a debugging framework for quick resolution of time-critical defects in proprietary customer runs taking days on hundreds of CPUs in a software system written by multiple developers with varying skill levels.• Filed 3 patents

Wilsonville, Or, Us

• Co-designed and implemented critical improvements to OPC product in simulation location placement and orientation, geometrical mask constraints checking, and language parsing. Supported time-critical introduction of the code into production environment. Coauthored a patent for “Site control for OPC” (7073162).• Implemented optical flare estimation using convolution, speeding up the original version 10x using lookup tables, multi-resolution simplification and multi-threading.

Wilsonville, Or, Us

• Designed and implemented GDS to VSB11 (Toshiba mask writer format) converter and VSB11 viewer interface, meeting aggressive performance constraints. Product was sold to Intel and DuPont Photomasks and is still used in production.• Improved polygons to trapezoids converter used in all vector fracturing formats by reducing shot count by 20%, adding rotation/mirror invariance and CD-cognizance.

Topica provides a web-based e-mail marketing automation service to online retailers, publishers, small and medium sized businesses, in addition to advertisement-supported hosted e-mail list services.At Topica, I designed, implemented and launched scalable, efficient, fault tolerant infrastructure in areas of search, advertising tools and database interactions for a website providing advertisement-supported email list hosting to 200 million subscribers. Contributed individually and as a lead of a 3-engineer team.• Implemented business logic in Tuxedo, database interactions in Pro*C, advertising tools using Unix daemons written and C. Used Perl XS to communicate with Apache mod_perl, perl to generate HTML pages. • Wrote MIME email message decoder used to strip non-binary data for indexing using Verity’s K2 search toolkit for millions of email messages.• Integrated public key encryption library for customer credit card information encryption.• Wrote complex SQL reporting queries. Designed schemas and indexes for the second-generation paid targeted advertisement service.

Researched, designed and implemented a SNR-scalable video codec based on Matching Pursuits algorithm achieving better quality than MPEG-4 and H.263+ at low bitrates.

Masters, Eecs

B.S., Eecs