Resume

Download PDF

ben@benashdown.com · github.com/bashd4 · U.S. Permanent Resident (ITAR U.S. Person) · Australian Citizen

Co-founded a YC startup (F26) that automated ~$15M in R&D tax claims — $100k ARR in five months. Four years at Google: data centre fabric design, then performance engineering for batch data processing spanning ~26% of Google-wide compute.

Work Experience

Jan 2026 – June 2026

Co-founder & CTO | ClaimKit (YC F26)

  • Grew to $100k ARR within 5 months of founding — shipped MVP and reached first revenue within 2 weeks
  • Automated the preparation of ~$15M in R&D tax claims for ~30 companies, including fellow YC startups
  • Cut claim preparation from weeks of consultant-led work to ~3 hours of customer effort, at 6% pricing vs traditional consultant fees
  • Personally designed and built the core product on a 3-person team — an orchestration system coordinating specialised agents for document analysis, R&D assessment, and claim writing, processing hundreds of thousands of pages of customer documentation
  • Solely owned all engineering for the company — MCP and third-party platform integrations, agent infrastructure, cloud infrastructure and CI/CD, cost-optimisation systems, and internal claim-assessment tooling

May 2025 – Dec 2025

Co-founder & CTO | Two early-stage startups

  • Explored two products — legacy-code modernisation tooling and AI project management for Slack — killed both on weak traction

Oct 2021 – May 2025

Software Engineer | Google — Flume Efficiency

  • Improved efficiency of Google’s generic batch processing library (Flume), with a focus on accelerator performance — the library accounts for roughly 26% of Google-wide CPU, 20% of RAM and 14% of Disk I/O
  • Created metrics to understand performance of the system on Google’s latest gen TPUs
  • Alleviated slow model loading onto TPUs, delivering multi-million-dollar efficiency savings to Google’s Offline Bulk Inference infrastructure
  • Discovered and profiled issues specifically related to Accelerated workloads, and undertook several projects to alleviate them, saving millions in compute costs
  • Discovered and fixed deep technical issues related to Flume running on Google’s internal job scheduling infrastructure and how signals are surfaced by internal VMs, recovering significant fleet-wide capacity

Software Engineer | Google — Fabric Design

  • Developed the software modelling infrastructure producing build blueprints and traffic simulations to validate every new Google data centre fabric design before physical build
  • Flexible Apollo NPI — modelled port-level connectivity for flexibly-sized optical circuit switching (OCS) layers, enabling data centres of varying size on shared infrastructure
  • With one other engineer, owned all fabric modelling for minicluster — Google’s small-footprint co-location data centres — including mesh bandwidth augmentation for high-bandwidth ML workloads and port allocations for a new Top of Rack switch variant to alleviate supply constraints

Dec 2020 – Feb 2021

Software Development Engineer Intern | Amazon

  • Built an event-driven notification system for events and maintenance campaigns on AWS (S3, SNS, SQS, DynamoDB), tracking the full notification lifecycle through to resolution
  • Drove design discussions across integrating teams and presented the system to ~30 engineers across sister teams

Jan 2020 – July 2020

Software Intern | Deswik

  • Development & maintenance of a .NET based windows application for mine management and scheduling

Jul 2019 – Dec 2019

Technology Intern | Accenture

  • Infrastructure & Environments team member
  • Produced options papers for secure and scalable hosting solutions and environment solutions that formed part of an organisational change into containerisation
  • Developed proof of concept applications for containerised development and developer environment infrastructure

Education

2017 – 2021

The University of Queensland | Bachelor of Engineering (Software)

  • Honours Class I — GPA: 6.6 / 7
  • Deans Commendation for Academic Excellence (×5) | Hawken Scholar

Projects

2020

Process Isolation in OpenBSD | C

  • Implemented toy “zoning” of processes in OpenBSD
  • Processes in different zones could not perform IPC
  • Processes in different zones from the user shell could not be seen using utilities such as ps

Jul 2018 – Oct 2018

Spaceport Ai Deep Learning Course | Python

  • Created a character-detection CNN, Dog breed detection CNN, Twitter sentiment analysis RNN

Jul 2018 – Nov 2018

Multiplayer Game | Java

  • Collaborated using Git & GitLab, SonarQube and Jenkins for CI
  • Worked on Powerups & Combat features, allowing players to collect, store and use powerups
  • Worked on multiplayer networking, focusing on coordinating UI elements across all players (lobby screen, main game, mini-map, player’s health)

May 2018 – June 2018

Frogger Clone | C

  • Built on an atmega324A and displayed on a red/green led matrix, created a fully functional Frogger clone with buttons/ Joystick/ keyboard/ Seven Segment Display