Joshua Johns

AI & Software Engineer I at PMG

Joshua Johns

Software engineer in Dallas.

Base
Dallas, TX
Class
Full stack
Degrees
B.S. + M.C.S.

Work

  1. 2026

    AI & Software Engineer I

    PMGJan. 2026 - PresentDallas, TX

    Ad creative automation for big brands

    Highlights
    • Shipped 90+ full stack features and fixes across a React, TypeScript, Redux Toolkit and RTK Query frontend and a Django and PostgreSQL backend for Creative Studio, an ad creative automation platform that replaced the legacy Dynamic+ toolchain for 25+ enterprise brands and cut new ad time to launch from 6 to 8 weeks down to 3 days.
    • Built the feature flag gating and on demand ephemeral preview environments that carried the product from internal alpha to general availability, backing the migration that moved 100% of the company's social catalog clients off legacy tooling before it was sunset.
    • Built the delivery destination layer that publishes finished creatives to Meta, Pinterest and Reddit, covering client level destination settings, a Manage Destinations UI that links templates to destinations, per destination variant and size selection, feed compatibility filtering on the link pickers, and automated product catalog syncing that replaced a manual spreadsheet and email handoff.
    • Engineered real time progress tracking, job notifications and a reusable unsaved changes safeguard adopted across the app for render and delivery jobs that fan out to 130K+ ad variants per run, and paginated a template list that previously rendered every record on load.
    • Cut load cost and lifted throughput across the platform by adding a template scoped render info endpoint and moving single template callers onto it, removing a creative count query no caller consumed, raising web concurrency, correcting an autoscaling trigger that was never firing, and releasing a user row lock that had been held for the full length of every request.
    • Traced and fixed silent data loss in the product feed pipeline, where uploaded files bypassed CSV sanitization and any row containing a newline dropped every column after it on read, then added the pytest coverage the command had been missing.
    • Migrated core modules to TypeScript with Zod runtime validation, moved the variant table and delivery sections onto React Hook Form, and normalized template sizing into a canonical relational schema.
    • Diagnosed a stacked staging failure that had kept the render dependent end to end suite red for days, using Datadog to separate an already fixed render 500 from the real cause, a screenshot renderer whose container had run out of disk so Chromium could no longer launch, leaving renders hanging and the delivery dead letter queue sitting at roughly 78,000 messages. Wrote it up with fix steps, and the gap it exposed led the team to add the staging monitors that would have caught it.
    • Handled Tier 2 production support for render, delivery and asset upload failures, reviewed frontend and backend pull requests across both Creative Studio repositories, contributed fixes upstream into the shared Alli design system, and onboarded a new engineer onto the codebase, template build workflow and release process.
    • Designed and built Channel Captain with two other engineers, a Python Slack bot that governs channel creation, naming enforcement and ownership across the company's enterprise Slack workspace, with modal and slash command workflows that validate the name, apply the correct prefix and normalize input before a channel is ever created, plus scheduled audits that flag stale channels and notify their owners.
    • Took Channel Captain through enterprise security review by cutting its requested Slack admin permissions from ten declared scopes down to the four it actually exercises, documenting the purpose of each for IT, and drove the approval and install into the production workspace. Backend runs on AWS EC2 provisioned with Terraform, with credentials in Secrets Manager.
  2. 2025

    Machine Learning Research Assistant

    ClassTranscribeMay 2025 - Aug. 2025Urbana, IL

    OCR and video pipelines so lectures are searchable

    Highlights
    • Built an end to end ML pipeline with OpenCV and Tesseract OCR to extract on screen text from lecture video, improving accessibility and search for 60,000+ students.
    • Designed a classifier for dual stream lecture recordings on ClassTranscribe, the University of Illinois lecture captioning platform, so slide content and speaker video are processed separately.
    • Engineered the video processing and summarization systems the pipeline runs on.
  3. 2025

    Graduate Teaching Assistant, Computer Science

    University of Illinois Urbana-ChampaignJan. 2025 - May 2025Champaign, IL

    Course staff for a class of 800

    Highlights
    • Led discussion sections, office hours and student support for a course of 800+ students.
    • Taught Python programming concepts to 50+ students in a weekly lab while managing 2 course assistants.
  4. 2024

    Cloud Engineer Intern

    KenvueMay 2024 - Aug. 2024Skillman, NJ

    Onboarding tools for Azure and AWS

    Highlights
    • Built NLP onboarding tools for Azure and AWS that cut cloud knowledge retrieval time 30% for 1,000+ employees.
    • Standardized AWS onboarding behind natural language interfaces so engineers could ask for what they needed instead of hunting through docs.
    • Assembled a React platform covering AWS, Azure and on premise estates, raising engagement with the internal tooling 25%.
  5. 2023

    Data Scientist Intern

    Johnson & JohnsonMay 2023 - Aug. 2023Skillman, NJ

    A chatbot and the pipelines behind it

    Highlights
    • Deployed a production Python and NLP chatbot that answered questions about the company's products.
    • Built scalable Python pipelines that aggregated and normalized labeling data across 2,000+ products.
    • Used NLP to group semantically similar phrases from product labels across multiple languages, so the marketing team could see which phrasing actually sold.
  6. 2022

    Resident Advisor

    University of Illinois Urbana-ChampaignAug. 2022 - Dec. 2025Champaign, IL

    On call for 500+ residents in the halls

    Highlights
    • Held a crisis response duty rotation covering the safety and wellbeing of more than 500 residents.
    • Worked with 20 resident advisors and 2 multicultural advocates to plan and run community programming across the hall.
  7. 2022

    Data Analyst Intern

    CaterpillarMay 2022 - Aug. 2022Akron, OH

    Automation and dashboards for the data team

    Highlights
    • Freed over 20% of employee time by replacing manual data collection with Power Automate scripts.
    • Built SSRS dashboards on SQL that tracked and monitored Caterpillar's enterprise databases.
    • Designed Power BI dashboards with trendlines that fed strategic planning.
  8. 2022

    Lead Full Stack Developer

    Reflections | ProjectionsMar. 2022 - Dec. 2022Urbana, IL

    Registration site for the Midwest's largest student run tech conference

    Highlights
    • Led MERN stack development of the registration and scheduling site for the largest student run technology conference in the Midwest.
    • Drew 10K+ unique visitors and converted them into 2,000+ event registrations.
    • Collected attendee resumes and distributed them to conference sponsors including Qualcomm, Visa and Intuit.
  9. 2021

    Software Engineer Intern

    Snap-onMay 2021 - Aug. 2021Richfield, OH

    Moved an internal site onto modern browsers

    Highlights
    • Migrated an internal website used by over 10% of the workforce so it ran correctly on modern browsers.
    • Set up Jenkins for that website so development teams could test new features before release.

Projects

  1. 01

    pinmix

    Shuffle any Pinterest board into a new board in one click

    2026TypeScriptChrome ExtensionMV3

    • A browser extension that reads the board you are looking at, creates the new board first so a taken name fails immediately, then loads, shuffles and saves every pin into place. The original board is never touched.
    • Runs inside pinterest.com on the session you are already signed in with, so there is no login of its own, no browser automation, no server and no analytics.
    • Sections come along too. Keep them and each one is recreated and shuffled on its own, or turn that off and every pin mixes together. An optional seed reproduces the same order later.
    • Zero runtime dependencies. Built with tsc, tested with the Node test runner, and released as a signed zip by a GitHub Actions workflow on every version tag.
  2. 02

    Finish Radar

    Weekly ranked construction leads for commercial cleaning companies

    2026Next.jsTypeScriptPostgreSQLDrizzleZodPlaywright

    • Every Monday a commercial cleaning company gets the Dallas Fort Worth construction projects worth calling this week, why each one fits, who to contact and when to reach out.
    • Ingests public construction and permit sources on a schedule, deduplicates records against stable source identities so a rerun never doubles the dataset, and scores every project through versioned exclusion and fit rules.
    • Strict TypeScript end to end with Zod validation at every external boundary, Drizzle migrations, Vitest unit coverage on the scoring rules, and Playwright end to end specs.
    • Sending, scheduled production jobs and live payments are disabled by default, so the pipeline can be rehearsed against real data without touching a real customer.
  3. 03

    Creator Dunning

    Failed payment recovery for creators on Stripe

    2026Next.jsSupabaseStripe ConnectResend

    • Creators connect their Stripe account once and the app watches for failed subscription charges, then runs a dunning sequence that wins the payment back before the subscription dies.
    • Full Stripe Connect OAuth handshake with a signed webhook endpoint for Connect events, a scheduled dunning route, and transactional email through Resend.
    • Supabase Postgres with row level security, SQL migrations for the schema, generated database types, and separate browser, server and admin clients so the service role key never reaches the client.
    • The marketing side ships with it too. A recovery loss calculator, landing pages for each platform, long form guides, plus the sitemap and robots routes.
  4. 04

    Parallel Terrain Pathfinding

    Parallel Floyd-Warshall over real elevation data

    2025C++GeoTIFFParallel Programming

    • A parallelized Floyd-Warshall implementation that finds optimal routes across 3D terrain, weighting each step by both horizontal distance and elevation change rather than distance alone.
    • Reads real topographical maps as GeoTIFF elevation rasters and turns them into a weighted graph, then solves all pairs shortest paths across it.
    • Ships a visualization pass that renders the terrain and the computed route, tested against a San Francisco elevation map.

Skills

Languages
Python, TypeScript, JavaScript, Java, C/C++, C#, SQL, HTML/CSS
Frontend
React, Redux Toolkit, RTK Query, React Router, React Hook Form, Zod, Material UI, Tailwind CSS, Angular, Next.js
Backend & APIs
Django, Node.js, Express.js, Flask, .NET, REST APIs, Celery, Redis, Slack API, Stripe
Data & ML
PostgreSQL, MySQL, MongoDB, Neo4j, Snowflake, pandas, NumPy, scikit-learn, spaCy, OpenCV, Tesseract OCR, Amazon Athena, Drizzle, Supabase
Cloud & DevOps
AWS EC2/ECS/S3/CloudFront, Azure, Google Cloud, Terraform, Docker, Jenkins, GitHub Actions, Linux, Datadog, Playwright, OpenFeature, Flagsmith, Vitest

About

I work on Creative Studio at PMG. It's the ad platform behind 25+ enterprise brands, where one template becomes 130,000 finished ads in a single run. I've shipped over 90 features across the React and TypeScript frontend and the Django and Postgres backend, and the platform took new ads from a six week launch down to three days.

My favorite thing I've built there is the delivery layer. It publishes finished ads straight to Meta, Pinterest and Reddit, keeps product catalogs in sync on its own, and tracks every job in real time as it fans out. I also built a Slack bot with two teammates that manages channels for the entire company. It runs on AWS with Terraform and passed enterprise security review.

Before PMG I did a five year B.S. and M.C.S. at Illinois and finished a semester early. I taught a class of 800 as a TA, interned at Kenvue, Johnson & Johnson, Caterpillar and Snap-on, and built an OCR pipeline that made lecture videos searchable for 60,000 students.

Outside of work I like building things end to end. A Chrome extension on the Web Store, a lead scoring pipeline, a Stripe tool that recovers failed payments. They're all up in the projects section.

University of Illinois Urbana-Champaign

  • Master of Computer Science, GPA 3.66. Jan. 2025 - Dec. 2025.
  • B.S. Computer Science, GPA 3.58. Aug. 2021 - Dec. 2025.
  • Five year B.S. and M.C.S. program, completed a semester early
  • Grainger College of Engineering, James Scholar, Dean's List

Coursework

Parallel Programming Models (CS 598 PPM), Applied Machine Learning (CS 441), Natural Language Processing (CS 447), Artificial Intelligence (CS 440), Programming Languages & Compilers (CS 421), Database Systems (CS 411), Algorithms & Models of Computation (CS 374), Data Structures & Algorithms (CS 225)

Certifications

  • SnowPro Associate: Platform Certification, Snowflake. Issued Apr. 2026, valid through Apr. 2028.

Contact