Custom software development services for early-stage startups.

Community of the most intimate, vetted founder community ❤️Where founders help founders build — from Stanford, YC, & more! 😮

Building an enterprise-grade Snowflake emulator, including query processing, data storage / retrieval / analysis, UDFs, integrations, and more: https://snowflake.localstack.cloud/

- Stealth B2B2C mortgagetech venture to help mortgage brokers increase win rates by automating the document collection process.- In 3 months, built out a digital twin for the Universal Residential Loan Application (1003) for (co-)borrowers. Leveraged Next.js, Nx, Supabase, and Flask.- Negotiated access to Plaid's original bank statements beta program.- Built and shipped a homeownership assistance AI 72 hours after OpenAI's beta Assistants product launch.

Selected to On Deck's 17th Founder Fellowship cohort, which provides in-person summits, online community, connections and events to help find co-founders, make early hires, build early product momentum, get distribution, and raise capital.

- Wrote, tested, and published bespoke Python generalist coding practicum for interviewing software engineers during onsites. Popularity tied at #1 w.r.t. candidate selection for all engineering.- Led effort to redefine company sales funnel in fact and dimension tables, leading initial team-wide dimensional modeling and lakehouse migration initiative.- In a cross-collaborative team, Integrated Oracle NetSuite to internal monolith to migrate accounting operations from manual data entry contractors to automated pipeline under emergency circumstances. Redefined company general ledger in SQL for use in event-driven architecture. Turned project from red to green for reporting to board of directors, query runs <1s for operational purposes, ledger fully reconciled and approved by finance after 8 query iterations.- Securely processed and sent data to credit bureaus to validate capital pool expansion initiatives, assisting in leading central company-wide initiative across multiple departments.- Scaffolded data ingestion effort for contact center telephony migration and contact center optimization.- Assisted in company-wide data pipeline outage and migration from Stitch Data to Fivetran / bespoke solution.- Working with data science on productionizing machine learning models, e.g. addressing security and compliance concerns.- Refactored data pipelines using extensible partitioning strategy to save 80% of our BigQuery Analysis costs, or $125,000 per year.

- As acting tech lead, reviewed IC design and implementation of single-touch marketing attribution, offline ads conversion platform hydration (enabling 3x marketing spend w/ only 10% increase in CAC), built data system health metrics for CTO/VP reporting (4 9's reliability), Airflow 1.x to 2.x migration, data quality testing/alerting from bespoke to Monte Carlo, dbt analytics layer integration, BI platform migration from Google Data Studio to Metabase.- Founded and scaled the data engineering team from myself to 10 people, including writing and publishing job descriptions, evaluating technical screening tools, developing and validating multiple bespoke coding practicums, interviewing >20 job candidates for {PM, EM, IC} roles, onboarding new team hires, mentoring/training junior ICs, drafting a $250k+ contract engagement, etc.- Addressed BigQuery internal Sev 1 outage by migrating all pipelines based on GCP DataFlow / Python to GCP Composer / BQ Standard SQL in 1 week, eliminating identical future Sev 1s, increasing IC SRE capacity by cutting back core data platforms from 3 to 2, and reducing data warehouse refresh latency from 6 hours to 10 minutes.- Set up a custom BigQuery reverse ETL pipeline, hydrating our then-new Braze (email/SMS marketing tool) datastore with 20+ customer attributes, and flattening marketing spend on Segment.- Developed Equifax and TransUnion credit report ingest and normalization pipeline to BigQuery, expanding industry-differentiating customer propensity modeling capabilities.

- Prototyped native Pulumi integration with a JavaScript monorepo framework for GCP deploys of event-driven architecture components- Created fact tables using BigQuery and templated into existing sales/marketing scheduled reports to improve data warehouse modularity and query implementation time.

Tooling / Infrastructure Exploration- Created personal dotfiles to formalize development processes: https://github.com/yingw787/dotfiles- Achieved beginner-level proficiency with Haskell: https://github.com/yingw787/thehaskellbook - Built TinyDevCRM, a hybrid OLTP/OLAP streaming BI framework composed of PostgreSQL, Apache Parquet, and HTTP/2 Server-Sent Events: https://docs.tinydevcrm.com- Started Paint Creek Software (software contracting), >$3k revenue in 2020 (built out a logo comparison PoC using React.js + AWS Mechanical Turk)

Business-critical queries, in seconds, on TBs of data.Product Owner, Kinetica I/O (Python-based Graph ETL orchestration tool for highly structured and semi structured data)- Integrated Apache Spark as a data acceleration layer to boost data ingestion speeds by >2000x (1K records in 4 seconds → 20M records in 35 seconds)- Implemented engineering processes to shorten sustainable feature development lifecycle from 3 weeks to 1 day, enabling multiple $1M - $10M+ enterprise B2B commitments.- Implemented support for ESRI Shapefile ingestion supporting GIS use cases within Kinetica, part of the core product offering’s market differentiation and value add.- Designed and implemented data-driven integration test harness in support of rapid test-driven development with support for test auto-generation, improving mean-time-to-repair by >3x.

Empowering American manufacturing through dynamic part pricing and analysis and machine shop job routing.Web Services | Research Engineering- In a team of two, architected and wrote initial library for business logic validation during API V1 -> V2 migration.- Migrated logging infrastructure from Logstash to AWS CloudWatch + AWS Lambda.- Refactored machine learning pipeline for pricing and win rate calculations, brought testing coverage to ~80% and integrated individual testing coverage reports with gated commits.- Worked in a team of three to refactor pricing algorithms, contributing to a quoting performance boost of ~80%.- Prototyped front-end expedited shipping feature, contributing to primary expedite functionality which generated >$10K in its first week in production.- Migrated sections of internal ERP software from Jinja / Angular / jQuery mix to React, boosting developer productivity by >5x and improving page performance by >3x.- Migrated landing page from WordPress to in-house solution by developing Craft CMS templates and bindings, allowing sales and marketing to add site content without developer input while scaling to 4,000+ requests per day.- Assisted sales and marketing team in adding new partner activity monitoring abilities through Hubspot, and operations team in migrating shipping APIs to Shippo.

Worked in AT&T Foundry to develop and prototype benchmarking tests for OpenStack on IBM POWER8 systems

Creating and sustaining Duke University's maker community.- Recruited board of directors, who raised >$10K from alumni donations for equipment, and ~40 club members.- Obtained >1600 sq. ft. makerspace and funds for two new 3D printers from Duke AIAA and Duke Robotics.- Raised $2500 to make 3D printed lanterns for Pratt Engineering's 75th Anniversary Gala.- Popularized and supported Duke Innovation Co-Lab's Innovation Studio: https://colab.duke.edu/studio.- Designed and prototyped 3D printed biosand filter concrete mold, enabling 1000x savings vs. machined steel molds.

Developed a driving simulator environment for Google X's autonomous car research to simulate and test boredom with drivers in autonomous vehicles

Worked with two graduate students to design a prototype for a new teaching method, based on computer vision instead of manual control, for industrial robots.