Tom Harper Tom Harper Tom Harper Tom Harper

Tom Harper

Santa Cruz, CA • San Francisco Bay Area

Principal Engineer & CTO · Agentic Commerce, AI & Real-Time Media

Principal Engineer and CTO who turns ideas into shipped products — building early-stage companies from 0 to 1, several reaching 100M+ customers. My focus is using AI to augment the human experience, and I'm currently working on agentic commerce. I've guided teams of 200+ and bring deep expertise in real-time media systems — text, audio, video, sensor data, computer vision, and AI/ML.

I've chartered and guided hallmark products end to end. In Amazon Last Mile I helped initiate and lead VAPR, Smart Glasses, Driver In-Dash Experiences, Edge Safety AI, and Location Understanding. At Alexa I drove the launch of Echo Auto and 100+ devices bringing the Alexa wake word to every customer, and I was a founding engineer on Alexa Communications, whose core frameworks are used by millions. My work spans hundreds of services and millions of devices across edge, web, and mobile.

I've managed teams of 40+, mentored hundreds of engineers, and helped hire 1,000+ as an Amazon Bar Raiser. 4 granted patents + 4 in process. External recognition includes TechCrunch Startup Battlefield finalist, PC World, PC Magazine, and Microsoft Product of the Year.

Featured Work & Press

Vision-Assisted Package Retrieval (VAPR)
AI-powered system using computer vision to help drivers locate packages.
Amazon Smart Glasses
AR glasses providing hands-free navigation and delivery assistance.
Rivian Delivery Vehicle Software
In-vehicle infotainment and delivery management system.
Amazon VAPR (2024): TechCrunch: Amazon's new AI-powered vision tech tells drivers which packages to deliverAmazon Blog
Amazon Smart Glasses (2024-2025): TechCrunch: Amazon unveils AI smart glasses for delivery driversAmazon Blog
Rivian Infotainment (2024): EVWorld: Rivian's Infotainment Revolution - AI at the Wheel, No Smartphone Required
Echo Auto (2019): TechCrunch: After over a million pre-orders, Amazon's Echo Auto has begun to ship
Echo Buds (2019): GeekWire: Amazon unveils new Alexa-powered Echo Buds, undercuts Apple's AirPods with $129.99 price tag
Xiaomi Switchable Wakeword (2019): India Today: Xiaomi Redmi Note 8 with Alexa switchable wakeword integration
Droidcon (2022): Story about how Amazon built their first in-vehicle delivery appTom Harper & Lingshuang Wu presentation
Mobcrush (2015): GamesBeat: Mobcrush launches mobile game streaming on Android
ShowKit (2014): TechCrunch Disrupt Battlefield: ShowKit - A Mayday button for any mobile device
Tuul (2014): Silicon Valley Business Journal: Tuul's bots and app take on customer serviceSanta Cruz Sentinel: Santa Cruz tech startup Tuul hiring developers

Technical Writing

The Complete Modern Voice / Multimodal AI Stack
End-to-end real-time conversational AI — transport, endpointing, multi-stream fusion on an event-time timeline, the LLM core, the latency budget, and the session/KV scheduler. A full reference for building voice agents.
How the Low-Level GPU KV-Cache Works
From the attention math up through PagedAttention, prefill vs. decode, swap-vs-recompute, and why barge-in cancellation is cheap. The systems view of LLM inference memory.
Voice Latency Design
Where mouth-to-mouth latency actually goes, why endpointing dominates, and how to mask the milliseconds you can't remove.
GPU Data Movement & Serialization
Serialization formats and the path data takes onto the GPU — safetensors, Arrow/Parquet, pinned memory, zero-copy, and when to reach for each.
Multimodal Fusion & Co-Sequencing
Aligning audio, vision, and text streams that arrive at different latencies onto a single event-time timeline.
LLM Inference Server — Batching & Disaggregation
How continuous batching and prefill/decode disaggregation work in a modern inference server.
DAG Task Scheduler — Topological Sort
Scheduling tasks with dependencies via Kahn's algorithm, visualized.

References & Cheat Sheets

Python + Redis Practical Cheat Sheet
The stdlib and concurrency primitives you reach for under time pressure, plus Redis patterns for distributed request routing.
Algorithm Complexity Cheat Sheet
Big-O for the common data structures and operations, organized the way it comes up in interviews.

Talks & Frameworks

Work-Out Sim — a Monte Carlo Idea Generator
Reimagining GE's Work-Out for the agentic era: grounded agents argue over your codebase to surface high-leverage friction, scored as a distribution rather than a single answer.
Engineering Work-Out for the Agentic Era
A framework for stripping bureaucracy and making same-day engineering decisions.  Work-Out process (PPTX) · Engineering alignment (PPTX)

Experience

Principal Engineer - Amazon (April 2026 - Present)
Building in agentic commerce. (Santa Cruz, CA · Hybrid)
Founder - BrandCapsule · Freelance (Feb 2026 - Apr 2026)
Independent venture.
Principal Engineer - Amazon (November 2015 - 2026)
Driver Assistance Technologies & Safety (Aug 2023 - 2026)
Working on Driver Assistance Technologies (DAT) and Driver Safety (DIS). Combining AI/ML strategies with custom and OTS hardware to solve difficult real world problems in driver safety and productivity. Focus on humans-first approach to augment human understanding and capabilities. Assisted on VAPR, Smart Glasses, Safety Alerts, Hazard Detection, Location Intelligence, and AI Enhanced Personal and Developer Productivity.
Last Mile Technologies (Sep 2020 - Aug 2023)
Working to scale delivery experience at 1, 3, and 5 year time scales. Led software for Rivian Electric Delivery Vehicles and driver safety systems utilizing machine learning and real-time data processing to prevent incidents.
Alexa Accessories & Communications (Mar 2019 - Sep 2020)
Enabled Alexa Accessories including Echo Auto (with switchable wakeword functionality), Echo Buds, and Frames. Focused on developer productivity and app performance, driving cold start times from awful to industry standard. Introduced architectural modularization to deliver cross-platform features 10x faster with equivalent quality.
Alexa Mobile & Communications (Nov 2015 - Mar 2019)
Software architect for initial release of audio/video/messaging for Alexa Communications. Led cross-platform mobile rendering, smart home systems, messaging and notifications, Voice SDK, and Alexa Accessory Kit development.
Co-founder & CTO - Tuul (May 2014 - Aug 2015)
Led team of 25+ embedding automation and bot responsiveness into text messaging for customer service and sales. Filed 2 patents. Architected scalable infrastructure based on NoSQL (Cassandra) and distributed processing frameworks.
Principal Software Engineer - Logitech/LifeSize (Nov 2008 - Dec 2012)
Built software used by 25 million customers. Made high definition mobile video conferencing possible. Helped provide one of the first consumer 1080p video conferencing experiences.
Lead Engineer - SightSpeed (Mar 2003 - Oct 2008)
Engineering and R&D for real-time video encoding and delivery. Multi-award winning video conferencing client (PC Magazine/PC World Best Communications Products, Codie & Frost & Sullivan awards). Acquired by Logitech in 2008.

Key Accomplishments

TechCrunch Disrupt Battlefield 2014 Finalist
Competed at one of the technology industry's premier startup competitions
8 Patents (4 Granted + 4 Pending)
Innovations in video encoding/decoding, messaging systems, workflow management, and remote device command initiation
Video Collaboration Pioneer
Over 10 years of expertise in real-time media systems. Part of initial Alexa Communications launch.

Patents & Innovation

System and method for implementing workflow management using messaging
Innovations in using messaging systems to enable workflow automation and management
System and method for managing electronic conversations
Advanced techniques for conversation management and threading in electronic communications
Methods and apparatus for encoding and decoding video data
Advanced techniques for efficient video compression and transmission
System and method for archiving messages
Innovations in message archival and retrieval systems
Remote initiation of commands for user devices
Smart device automation and remote control systems

Technical Expertise

Cross Platform Development: iOS, Android, Windows, OSX, Linux

Compiled Languages: C, C++, Objective-C, Swift, Java, Kotlin

Scripting: Python, Matlab, Javascript, Typescript

Databases: Cassandra, MongoDB, MySQL, Oracle, PostgreSQL, Neo4j, Redis

Distributed Systems & Graphics: Shaders, OpenGL+ES, Metal

Signaling & Messaging: SIP, XMPP, Proprietary protocols

Streaming & Media Transport: RTP, RTMP, HLS

Audio & Video Codecs: H.263, H.264, H.265, VPX, Opus

Computer Vision & ML: PyTorch, VLM, LLM

Top Skills

Education & Professional Memberships

UCLA (1989-1995)

Professional Memberships:

Languages