Brian Bien – San Francisco, CA 94103 – email@example.com
Entrepreneurial innovator specializing in data-driven insights for problems like realtime bidding, ROI prediction and resource forecasting with ML. Production experience with machine learning tools and techniques using scikit-learn, Python and SQL.
AdWords Solutions Engineer (2017 – Google) – Data-driven Google Ads solutions – using ML to discover and explain best practices in the automotive industry.
Technical Account Manager (2016 – 2017 – Google) – Scalable digital advertising with DoubleClick and Ad Exchange. Solving publisher technical support and applying machine learning to our business strategies with scikit-learn.
Search Marketing Software Development (2008 – 2013 – BienTek, LLC) – Built a unique onsite contractor services business. This business is highly automated using a custom-written, PHP-based CRM tool with integrated PPC bid computation, contractor ranking, and job dispatch features. BienTek was sold in 2013.
Local Coupon Search Website (2009-2010 – SaveNextDoor, LLC) – Co-created, marketed, and auctioned a local coupon browsing site. As owner and co-founder, I defined site specifications for the developers, focusing on SEO, site architecture, and internal ranking algorithms.
AdWords Solution Engineer
As an AdWords Solution Engineer, my goal is to find opportunity in data, then translate this into concrete steps that help both the customer and Google perform their best. AdWords customers are encouraged to apply best practices to their campaigns, but what does the data say about the importance of these best practices? Machine learning helps to answer this question, yielding next steps and focus areas for our customers.
Technical Account Manager
Focus: identifying opportunities to scale Google’s global support capacity with machine learning and forecasting.
To address capacity planning, I proposed and implemented a forecasting model to predict different outcomes related to the clients we support. This time-series analysis uses ML techniques to predict ROI. Different modeling approaches are combined to reduce uncertainty.
Experimental projects include project similarity matching and case-agent assignment prediction for customer support automation.
My current focus is on modeling cost drivers for our department to produce explainable models. Emphasis is on transforming and visualizing data in novel ways to bring about original insights into costs and trends, allowing us to prepare for changes ahead.
A recurring theme of this program’s earlier phase was to spot an anomaly, investigate the data, and formulate and test hypotheses about what went wrong prior to coding a solution. The primary related, immediate challenge was improving and automating the data synchronization processes, which rely on feeds pulled from Audi, as well as other internal programs on our network. SQL queries were constructed to better understand the data feeds that the system relies on. An ongoing challenge was to make inferences about optimal business logic by observing changes in the data over time, and to update the system accordingly.
Since my arrival, system automation had been greatly enhanced by replacing manual data entry processes with feed-based automated updates (using PHP and MySQL). Processes that normally took weeks were reduced to days; manual inspections of data were replaced by heuristics to alert us when updates are necessary. We have gained insight into process trends and relationships using interactive Google charts to create a clear big picture internally, allowing us to speak intelligently with Audi when proposing product updates and new features.
To ensure system stability and data integrity, a custom event notification system was implemented on my initiative. This system catches errors, anomalies, and performance metrics to integrate into a unified report of activity. These disparate sources of information are processed and categorized to notify the admin on an urgency-determined basis; in this way, we balance vigilance and information awareness with a minimization of interruption frequency.
Since my arrival, significant key performance indicators have shown desirable trends: email cancellations – often an indicator that something is wrong – have dropped significantly over time (18% in 2013, 7% in 2014, and under 5% in 2015). Despite system growth, the average page load time has decreased from 3 seconds to under 0.5 seconds. Meanwhile, QA time has continued to decrease – much of which can be attributed to automated updates, the introduction of unit tests, and the event notification system mentioned above.
As the former owner of a scalable, data-driven onsite IT services business with over 10 active contractors – my primary roles included software developer, data analyst, and project manager. Daily tasks included PHP & MySQL software development, hiring, sales, accounting, search marketing, and business strategy.
At the core of BienTek is a custom CRM program with integrated pay-per-click bid optimization features. This software is used to automate business processes like contractor management, ad campaign management, and keyword bid calculation for AdWords and Bing Ads. Performing the roles of project manager and software developer, I lead the development of BienTek’s CRM software. Approximately half of this tool’s code was delegated to another software developer; I personally coded the back-end (focusing on ranking algorithms, keyword bidding, expected value calculations, and lead tracking). The PHP-based CodeIgniter framework was selected for this project in order to facilitate web accessibility while focusing development efforts on key tasks like ROI tracking and contractor management. This web-based software uses a MySQL database to associate data from job profits with contractors, skills, leads, and keywords; using this data, BienTek achieves online advertising efficiencies that permit a pure subcontracting business model to exist. Data tracked in the program also drives a proprietary contractor ranking algorithm for job assignments, helping to statistically predict which contractors are most likely to succeed with a job. Technicians use this web-based software to accept jobs, report on status, record payments, and invoice BienTek.
In my search engine marketing role, I ensured that the business performed well for popular organic search keywords, achieving the first position in Google for queries like “Michigan Computer Repair” (despite having no physical storefront.) To maximize paid advertising ROI, thousands of keywords were bid on in AdWords and Bing Ads for combinations of geography, time of day, and service type, according to changing contractor availability for different regions. Ad Groups and their Ads were dynamically generated along with dynamic tracking URLs for system performance feedback. A WordPress blog and various microsites supported BienTek’s presence in the organic SERPS.
BienTek required significant remote project management, much of which was performed through Upwork (formerly “oDesk”), an online recruiting and outsourcing platform. In the employer role on this jobs marketplace, I have outsourced projects for BienTek and SaveNextDoor ranging from software development to research, earning a feedback rating of 4.99 / 5 over the span of 53 project reviews (2,245 hours billed by my remote service providers). Through this remote project management experience, I created processes and documentation for contractors; qualified candidates through testing; and defined software specifications for website development.
The sales and networking aspects of BienTek included cold-calling and meeting with contractors, combined with the online promotion of a landing page for IT contractors seeking jobs to generate inbound inquiries. Technicians were interviewed and background-checked before becoming accepted as authorized IT contractors. Sales work included responding to incoming new project inquiries and calling technicians, pitching the opportunity to become one of BienTek’s trusted subcontractors.
BienTek was sold in 2013.
SaveNextDoor was a local coupon search business that was launched as an exciting way to get into web development while learning to apply SEO to a custom content management system. Our site architecture and content management strategy addressed challenging SEO concerns faced by most public CMS’s (e.g. duplicate content, short-lived content, content generation, and ranking content of different value).
My roles in this website included business planning and analysis, marketing, and definition of the site’s technical specifications like page architecture and ranking algorithms for implementation by co-founding developers.
In this business, a major challenge was getting business owners to invest time in creating coupons; free entry was not enough. It became apparent that solving this apparent catch-22 required either a managed data entry process, or sourcing coupons from feeds. The site was auctioned.
At Blackbaud – a company that produces fundraising software – I became familiar with CRM software and its application to marketing, lead generation, and lead retention. While there, I participated in the company’s CRM and SQL training, developed source code monitoring software using .NET to generate and store metrics; created unit tests; and programmed code fixes for Blackbaud’s flagship CRM product. I left Blackbaud to pursue my own business.
University of Notre Dame
B.S. Computer Science, University of Notre Dame (2002-2006)
> Received educational funding from scholarships and awards: Comcast Leaders of Tomorrow, Notre Dame Club of Detroit Scholarship, Lakes Area Chamber of Commerce Scholarship, Kondur Memorial Foundation Scholarship.
Bag of Words
This section includes some additional skills not mentioned above for the benefit of your resume relevance algorithm:
Skills: gradient boosting regressor, gradient boosting classifier, parallel computing, bayesian statistics, pandas, numpy, matplotlib, seaborn, time series forecasting, deep learning, data munge, unstructured data, unsupervised learning, version control, git, svn, linear modeling, support vector machine, boosting, bagging, data analysis, decision tree modeling, classification, quantitative research, quantitative marketing, machine learning, scikit-learn, global teams and processes, agile development, naive bayes, ensemble modeling, tf-idf, grid search, cross-validation, k-means, AI, engineering, partial dependence, prescriptive analytics.