Resume PDF

Bradley Leatherwood B.E.L

You bring the problem. I design the path and ship it.

Founder of inKibra, an AI development & deployment company. Former Google Search engineer — features used by a billion people. Named inventor on a granted U.S. patent. Fifteen years across startups, scale-ups, and Google, owning the work from product design through production.

Ex-Google Search — features reaching 100M+ users daily and a billion monthlyGranted U.S. patent US12602426B2 — named inventor and technical authorFounder/CEO of inKibra — AI services company earning up to $50K/month, owning product and engineering end to end15+ years across fintech, ad tech, automotive, entertainment, and search

Capabilities — toggle any combination to trace them through the career

Product Design & Management: product strategy · discovery · UX and product design · rapid prototyping (Claude · Figma · Framer) · turning ambiguous business problems into scoped roadmaps · requirements shaping · analytics-driven iteration · stakeholder managementAI & Data: OpenAI APIs · RAG · LangChain · agent workflows · eval loops · embeddings · TensorFlow · document processing · browser automation · analytics pipelines · audio/music analysis · HealthKit/Apple Watch context systemsGeneral Engineering: TypeScript/JavaScript · Rust · Swift · Python · Java · C++ · C# · React/React Native · Next.js · Expo · Node.js · Bun · Tauri · HTML/CSSCloud & Infrastructure: GCP · AWS · Azure · Cloudflare + R2 · Kubernetes · Terraform · Docker · Consul · Vault · Postgres · MySQL · PlanetScale · Redis · MongoDB · Couchbase · Cassandra · Elasticsearch · CrateDB · CI/CD · Sentry · PostHogReal-Time Audio: Web Audio API · AVAudioEngine · real-time DSP for music recomposition

Experience — flip a card to see the work

2022 — Present

inKibra, Inc.

Founder / CEO / CTO

One company, two tracks: client services on embedded hourly retainers, and a product lab that produced a granted patent.

Track 01 · Client Services

AI Development & Deployment for Businesses

Consulting · fractional product & engineering leadership

  • Own the full engagement lifecycle for every inKibra client — discovery, proposals and estimates, architecture, implementation, QA, release planning, and stakeholder communication.
  • Operate as product designer and manager as much as engineer: clients typically arrive with a business problem, a feature request, or raw user feedback rather than a spec — I design the specific path forward, from discovery and UX through architecture, delivery, and analytics-driven iteration.
  • Run an AI-first delivery process: inKibra's concurrent, multi-threaded AI development harness (open-sourcing soon) takes a described goal — a new feature, a latency target, a bug — and plans, builds, and verifies the work against explicit rubrics, with full visibility into what was done and how it was judged. Now being integrated into the construction client's own development process.
  • Serve as lead engineering partner for one of the largest construction firms in the Pacific Northwest — stadiums, hospitals, and nine-figure contracts — building roughly half of the firm's internal software, the systems connecting back office, front office, and field contractors.
  • Built the firm's internal PDF and document-management platform across web, mobile, and backend: document processing with revision tracking, validation workflows, and tenant-scoped safety-checklist normalization.
  • Built data-cleaning and reinforcement-learning data pipelines for a provider in Microsoft's data ecosystem.
  • Raised the organization's engineering discipline — PostHog feature flags, launch observability that automatically captures the context to judge each feature's success, Sentry, CI/CD, and preview environments — driving company-workflow errors to near zero through multiple mobile-app relaunches.
  • Deploy and operate production web, AI, and media workloads on GCP, AWS, and Cloudflare with Kubernetes, Docker, signed-URL object storage, and automated CI/CD.
ProductAI & DataEngineeringCloud

The flip side · how an engagement runs

From “we have a problem” to shipped, measured software.
▲ workflow errors: high

“Crews are double-entering safety checklists, and we can't tell which version of a drawing is current.”

No spec needed — I embed and find the real requirements.

stakeholder interviewsfield-crew shadowingconstraint: offline-first sitesUX flows draftedsuccess metric agreed

A system design that fits how the org works — buy-in when it matters; usually it's high-trust: bring the problem, I fix it.

Web · back officeAPI · documentsMobile · field crews

Shipped behind a feature flag — rolled out crew by crew.

checklist-v2 enabled

rollout 10% → 100% · preview env → production

▼ workflow errors → near zero

Observability judged the launch — then we picked the next problem. The solutions compound.

embedded hourly retainer ~½ the firm's internal software shipped this way

Track 02 · Product Lab

Recordless & ToneTempo

Inventor & product owner — audio AI, fitness coaching

  • Invented and patented interactive music recomposition technology (U.S. Patent US12602426B2) enabling live, user-driven rearrangement of tracks and workout-aware playback; raised angel funding to build it.
  • Engineered the patented mix engine in Rust, Web Audio API, and AVAudioEngine — BPM-aware transitions, equal-power crossfades, section-aware rearrangement, stem- and effect-aware mixing, and offline rendering — sustaining sub-30 ms recomposition-to-output latency.
  • Built an agentic AI fitness coach on OpenAI models with retrieval-augmented generation over workout context, nutrition guidance, and Apple HealthKit / Apple Watch data — personalized workouts with real-time coaching and motivation.
  • Architected and trained a deep music autoencoder in TensorFlow, reducing raw audio feature dimensionality by more than 90% and auto-labeling 1M+ "interactivity points" for directed recomposition.
  • Containerized and operated all web and AI services on GCP/Kubernetes at 99.9% uptime, load-tested to sustain 10,000 requests/sec.
  • Shipped ToneTempo to iOS and web — TypeScript, React, Tauri iOS, Swift plugins, HealthKit, Stripe and Apple in-app purchase — reaching thousands of users and ten onboarded creators through TestFlight alphas and production releases.
  • Owned product end to end — strategy, UX and product design, roadmap, and prioritization — alongside engineering, fundraising, marketing, creator-program design, and gym-pilot strategy.
ProductAI & DataEngineeringAudio

The flip side · inside the product lab

Songs that restructure themselves around you.

The engine labels a track's sections and recomposes them live — at sub-30 ms latency:

FIG. 2 — recomposition apparatus · U.S. Patent US12602426B2

102

source track

104

recomposition engine

BPM-aware · equal-power crossfades · stem-aware

106

adaptive output

<30 ms latency

IN
VERSE
CHORUS
CHORUS
VERSE
DROP
OUT
the original arrangement …duplicate the chorus …insert an extra verse …stretch sections to match your pace
An AI you talk with, not at.

Most AI chat is turn-taking: you wait. ToneTempo's coach attends to multiple threads in one chat — answering the quick question while a longer reply is still being composed.

↳ workout planRebuild today's workout — longer cooldown, new music.
composing your plan…
↳ quick questionQuick one — good warm-up BPM?
↳ answered instantly100–115 BPM — I'll start you at 108.
↳ workout plan · doneDone — cooldown extended to 8 min, playlist recomposed to land on slower sections.
Shipped, on the App Store, in use.

TypeScript, React, Tauri iOS, Swift plugins, HealthKit, Stripe and Apple IAP — through TestFlight alphas to production.

1000s

of users

10

creators onboarded

iOS + web

live surfaces

2020 – 2022

Google

Senior Software Engineer, Search

  • Launched local-answer action buttons (call, directions, food ordering) on Search — a 5% DAU increase on the surface, used by 100M+ people daily.
  • Rebuilt local text-to-speech in Geo on a more robust engine, cutting error rates 3% for a feature serving one billion users monthly.
  • Improved local-search accessibility through a framework-level change, raising accessibility scores for a billion monthly users while avoiding a far larger refactor.
  • Built an NLP speech-recovery module — "did you mean" recovery for misunderstood voice queries — that began as a hackathon win, unblocking the owning team, who shipped the final solution.
  • Pioneered a journey planner for Maps — plan a multi-stop itinerary (errands, dinner, a movie) around your route — a second internal hackathon win — and has since launched publicly.
  • Pioneered taste-based local search with shareable collections as a search overlay, matching venues through an array of ML methods including a proprietary GAN — a third hackathon win that catalyzed additional headcount and director-level OKRs.
ProductEngineering

The flip side · what shipping at Google scale means

Small buttons. A billion people.

1B

monthly users of Geo TTS + accessibility improvements

100M+

daily users of the local-answer actions

+5%

DAU lift from the launched buttons

hackathon wins — one since launched publicly

before

Tony's Little Sicily

4.6 ★★★★★ · Italian · 1.2 mi

after — launched 2021 · 100M+ daily · +5% DAU

Tony's Little Sicily

4.6 ★★★★★ · Italian · 1.2 mi · Open until 10 PM

📞 Call➤ Directions🍝 Order

before — brittle TTS engine

🔊 “Turn left on M—” ✗ playback error

after — rebuilt engine · −3% errors · 1B users/mo

🔊 “Turn left on Main Street.”

before

“navigate to piza nesr me”

“Sorry, I didn't understand.”

after — NLP recovery · shipped by the owning team

“navigate to piza nesr me”

“Did you mean: pizza near me?” Yes

before — one destination at a time

📍 Mall

after — multi-stop itineraries · launched publicly

🏬 Mall 🍝 Dinner 🎬 Movie

before — framework blocked screen-reader fixes

screen reader: “button”

after — highest accessibility score · 1B users/mo

screen reader: “Order food — button”

before — solo, one-shot searches

“restaurants near me”

after — shareable collections + GAN venue matching

Date night ideas

shared collection · 12 places · matched to your taste

+ Add place👥 Shared

3× hackathon wins · taste-based search drove new headcount + director-level OKRs

Luxe / Volvo

2018 – 2020

Senior Full Stack Developer

On-demand valet startup acquired by Volvo and rebuilt as the testbed for Volvo's consumer services in the run-up to the Polestar launch.

  • Joined immediately after the acquisition as one of three backend engineers turning the Luxe car-movement engine into Volvo service infrastructure — re-architecting backend workflows so the startup's valet engine could power Volvo's consumer service flows, the service foundation for the Polestar launch.
  • Led the SDK-first refactor (TypeScript AST codemods) that drew a clean API boundary between the backend and the five-person frontend team — unlocking parallel frontend development, eliminating duplicated code and cyclic dependencies, and introducing a Cap'n Proto / GraphQL–inspired pattern for bundling API calls.
  • Built the localization workflow (GitHub Actions, Docker) used by the Luxe and Volvo On Call teams; advised European GDPR and Chinese data-compliance architecture.
EngineeringCloud

MSMCI / Megastar

2017 – 2018

Engineering Team Lead & Senior Full Stack Developer

Entertainment startup creating live talent-show products with celebrity judges such as Usher.

  • Architected the microservice framework, self-documenting protobuf-inspired RPC tooling, and WebRTC-based sub-second event streaming that let live polls and chat games track the broadcast in real time — the show's engagement engine.
  • Led a six-person engineering team across React Native apps, release workflows, and FFMPEG/CDN/Wowza video pipelines on AWS with Terraform, Docker, and Consul.
  • Owned hiring — designed a methodology of phone screens, coding challenges, and team interviews, and grew the team by two engineers.
  • Helped set up and produce the end-of-season live show streamed by Usher — including firefighting a last-minute streaming-vendor settings change as the broadcast went live.
  • Led product development of a machine-learning social-media usage optimizer for brands and influencers — interviewed external stakeholders to identify market needs and opportunity size, collaborated on the business plan, and presented recommendations to senior management.
EngineeringCloudProduct

Trunomi

2015 – 2017

Lead Front-End Developer & Full Stack Developer

FinTech startup providing consent and customer-data management for banks and any company facing GDPR-era regulation.

  • Architected TypeScript microservice APIs with ledger-based data models, generated clients and SDKs, JWT authentication, and enterprise partner integrations.
  • Built a finite-state-machine tester proving complete state-transition coverage; owned cloud-security — managing a PKI and secrets with Vault and Keybase and setting the developer toolchain and best practices.
  • Contributed to the product roadmap with a component-based workflow-composition UI that sidestepped rules-engine complexity entirely — ASN.1/PKCS#7 for security, multi-part messages for legacy compatibility and extensibility.
EngineeringCloudProduct

Smaato

2014 – 2015

Manager, Sales Engineering

German mobile ad exchange with 80,000+ publishers and 500 demand-side partners.

  • Built the team's log-search and analytics tools — shared searches and saved analysis scripts that let the six-to-eight-person sales-engineering team cut through exchange logs to spot failures and usage patterns across billions of daily impressions — making key queries 10x faster.
  • Reported to the COO and Director of Global Sales Engineering; integrated demand-side platforms and ad networks into the exchange.
EngineeringCloudProduct

Independent Projects & Consulting

2013 – 2020

Consultant & Open-Source Author

Distributed systems, blockchain infrastructure, ad tech, and media consulting alongside full-time roles.

  • Authored EIP-1080, among the first formal descriptions of recoverable tokens and wallets — engaging the Ethereum community and the Kleros arbitration network on game-theoretic griefing constraints; the ideas fed later research and product categories like social-recovery wallets (e.g., Argent).
  • Led Horizon Games' scaling research and early prototypes — account-based Plasma, Tendermint-style consensus, a TypeScript/WebAssembly sidechain with a two-way Ethereum bridge — work that fed the core technology that became Sequence, the wallet platform now powering millions of onchain interactions.
  • Consulted for Bitfinex/Ethfinex on market-making incentives — built the reference solution on the 0x protocol with token auctions and a fee-collecting ERC-20.
  • Audited the designs that preceded Bitfinex's production Nectar token — eliminating reentrancy and overflow attacks, with simulation mocks to tune parameters against griefing and sybil attacks.
  • Built Smaato's first self-service ad-tag creation flow (React/Node, designed end to end) so clients could configure tags for any SDK without sales-engineer help — plus malicious rich-media testing pipelines, OpenRTB bidder integrations, and VAST compliance automation.
  • Migrated MSMCI's AWS infrastructure to a new account within 24 hours with zero app downtime, using Terraform, Consul, and Docker.
EngineeringCloud

Portal Games

2008 – 2011

Founder

Independent game studio and publisher — the first founder story.

  • Built a 3D game engine for Xbox 360 in C#/XNA — multithreaded physics, AI, rendering, and FPS mechanics — and released it open source.
  • Developed latency-aware networking that scaled server-client communication with the visibility and proximity of players' avatars.
  • Published games on the Xbox marketplace on behalf of other developers.
EngineeringProduct

How I Work

01

Problem

You bring a business problem, a feature request, or raw user feedback — rarely a spec. That's expected.

02

Discovery

I embed in your organization, dig into requirements and constraints, and design the product that fits.

03

Design

A system design that fits how your organization works — I get your buy-in when it matters; most work runs high-trust.

04

Build & ship

Web, mobile, backend — QA, release planning, preview environments.

05

Measure & improve

Feature flags and observability judge each release — then we improve, maintain, and start the next solution.

Engagements run on an hourly retainer: I embed inside your organization — understanding the problems, drafting solutions, then building, shipping, and measuring them. The work is one senior engineer multiplied by inKibra's AI development harness: describe a goal — a feature, a latency target, a bug — and the harness plans, builds, and verifies it against explicit rubrics, with full visibility. Over time the solutions compound, each one building on the last.

Book an intro call →

Education

2011 – 2014 Jacobs University Bremen B.Sc., Information Management and Systems — Operations management, computer science (operating systems, databases), and information systems
2012 Stanford University Operations Management & Client-Side Web Programming

Honors & Hobbies

HonorsJacobs University Student President · Jacobs University Scholar · National Merit Scholarship Finalist
HobbiesFrench (conversational) · Private pilot · Scuba diver · Ultimate frisbee