First DS hire to support the Infrastructure pillar at Figma.- Cloud efficiency and usage forecast- Cloud cost attribution and observability- Help discover insights for cost optimization- Bringing more data-decision decisions to the Infra pillarAI Cost data foundation

"How are we doing in messaging abuse from the member's perspective?"Created a measurement framework and strategy to better measure the member experience of messaging abuse on the platform. Successfully aligned the proposed framework across Product, AI, and Infra."Which anti-scraping mechanism is the best for LinkedIn?" - Designed experiments to test anti-scraping defense mechanism and used insights to drive product roadmap. - Worked with Growth to ensure proper guardrail on SEO and Growth metrics; steadily and gradually steering the ship forward."How do we protect members from emerging scam message?" - Built a model that uses TF-IDF to detect trending spam keywords - Pushed the model into production; queuing flagged messages to review teamUp-level the team - Held weekly office hours to help team mates strategize, prepare, and whiteboard various problems ranging from communication to ideation. - I rebranded a weekly team sharing meeting into a learning zone to help team members get reps in for upcoming presentations and deliverables. This meeting went from low signups and frequent cancellations to full signups and zero cancellation for the next 3 quarters.

Led the Market Intelligence DS team at Wish. - Spearheaded various initiatives to help connect the dots between market trends, pricing behaviors, and what Wish can offer to our users. - Led the development of a price check engine, which comprises a combination of vetted heuristics and Deep Learning models, to enable the large-scale competitive analysis in both pricing and supply. - Initiated potential project partnerships with various teams across Content, Business Development, Marketplace, and Relevancy. - Leveraged existing ML models to help various teams to achieve their business goals (faster first sale for new merchants, boosted search impressions, directing more sales to high-quality merchants, etc) - Mentored interns and junior Data Scientists - currently holds the record in number of interns mentored across the entire DSE organization at Wish.

As a Senior Data Scientist, I worked on... - A financial product that involves early payment to merchants - Data Modeling and Management - Python-based ETL (Luigi, Airflow) - Transaction risk modeling - Fraud detection with machine learning, various analytics, and more - User profile/score analytics, metrics generation - Logistics & inventory optimizationStuff I do on the side - Mentor interns and new hires - Conduct training sessions in onboarding sessions (biweekly) and intern data camps (~3x a year) - Building out the data team through interviews, career fairs, etc.

Data Engineering for Google Consumer Hardwares

I analyze large amount of user traffic data to obtain trends and insights to make more data-driven and intelligent decisions in capacity planning. To get my job done, I use Python, R, SQL to assemble, cleanse, and archive large amount of data from various sources to enable in-depth analyses. I also act as an Interface between Capacity Planning and Software Engineers to build data pipelines and improve data quality. I have deployed several curated data into Google production environment to enable consistent access and analysis of data.I also build both analytical and simulative models for traffic demand forecast.Some projects I'd worked on (or still working on):Built a user-interactive model that supported the implementation of increasing the maximum number of customers on a passive optical network from 16 to 32. Resulted in nearly 50% reduction in capacity augmentation cost.Built a simulation model to estimate network bandwidth peaks given the number of users on a network. This model is being used to drive business decisions on new Alphabet Access product offering.Developed a TV channel ranking metric based on viewed time and number of viewers per hour. The results were used by Video Ops team to determine optimal maintenance schedule and in negotiations with content providers.Joined demographic data with network traffic data to mine for correlations and patterns in user traffic behavior. Used algorithms like k-mean clustering and principal component analysis to analyze the data. The insights gained from this analysis were used to update and enhance quarterly capacity plan.

Serves as the main point of contact for the onboarding of new compute and storage product to the cloud infrastructure pricing program at Google; currently owning the pricing program for Accelerators and Mini-Clusters.Developed pricing models that reflect true operating cost through conducting in-depth analyses and data gathering; reduced a days-long process to minutes by automating the price generation process. [Google Sheets & GAS]Built a range forecast model for future server CPI using the Monte Carlo Method; modeled the variability of key inputs using probability distributions based on historical data, SME inputs, variability of similar metric, etc. [Python ~1000 lines]Developed a long-term headcount forecast model for Datacenter Hardware Ops team; $120M+ of annual operating impact. [Google Sheets, SQL]Designed, built, and delivered a risk model prototype that uses cumulating historical data to identify potential deployment slips for future projects; reduced buffer at a region by 2 weeks based on the results. [Google Sheets & GAS, R]Implemented a newsvendor-based inventory model that quantifies the benefits from achieving potential lead time reductions and flexible upsides; used by Managers in negotiation efforts with suppliers. [Google Sheets]

Gathered, cleansed, and analyzed large datasets on Google’s server part transactions, created inventory visibility for spare parts, and recommended solution with improved forecasting and inventory strategy. [SQL]Implemented the spares planning model in R (>200 lines); created data tables (in R) that serve as essential building blocks ($300M+ impact on annual operating cost) for a forward looking view of the Google fleet install base. [R, SQL]Implemented a DRAM price forecast model based on Log-Linear Mean Reverting method; this model was adopted by Resource Pricing Team, and later being used in server range forecast model. [R, Google Sheets, SQL]Developed a Consumer Price Index model to forecast future fleet machine cost; this model was later adopted by the Finance partners for budget calibration. [Google Sheets, SQL]

Statistics 2: Introduction to Statistics

Analytic Decision Modeling Using Spreadsheets

Oversaw the development of the internal metrics reporting dashboard – platform research, data hierarchy design, data field property design, functionality and user interface.Built a Burn Rate Model (Excel) that visualizes computing resource consumption rates derived from dynamic data.Performed ad-hoc reports and analysis on computing resource usage data with an internal language similar to SQL.Collaborated cross-functionally (SWEs/PM) to design and create new KPIs based on requirements from Director/PM.Assisted Team Liquidity with compute-budget initiatives such as inorganic growth forecasting, budget planning, price re-indexing and network chargeback implementation. Delivered a weekly report and presentation on the project status.

Designed and implemented Excel applications to automate cleaning and to format standardization processes for raw data.Assisted Haas MBA directors in providing accurate periodic dashboard reports on student salary and offer statistics. Provided business school ranking metrics for magazines like US News and World Report and BusinessWeek.Monitored the overall integrity of the database for job offers received by MBA students at Haas.

Created over 40 pages of functional and technical design documents for various software development projects.Worked cross-functionally within Cisco IT to ¬assist in creating network solutions for external and internal clients of Cisco.Participated in pilot programs for new technologies and debriefed to various Cisco engineering teams in India and USA.