Available Q3 2026 · Bay Area / Remote

Operations and data, from a robotics floor to a five-car fleet.

Data Operations Lead on Tesla's Optimus program. I keep a 24/7 collection pipeline running and ship the dashboards my team works off of. Outside Tesla, I operate a five-vehicle commercial fleet as my own P&L lab. Open to fleet operations, operations analyst, and TPM roles, especially at autonomous-systems and logistics companies.

Live fleet
M3P
M3SR · driving
MY
Volt
Yaris · driving
Fleet OS Agent / live demo
online
>
Try:
Personal fleet · operating
5 vehicles

Run as my own commercial P&L for 26 months. Live telemetry, real booking data, real maintenance decisions. Same playbook a fleet ops team uses, at one-thousandth the scale.

Day jobTesla Optimus · Data Ops Lead
ShippingFleet OS · DayFlow · agent demos
Open toFleet ops · ops analyst · TPM

Featured work

01 / SELECTED
Featured · Analyst loop

Anatomy of one fleet decision.

One vehicle. A 5-percentage-point utilization gap. A 30-day test. A keep-or-sell decision with a Q3 calendar and a $14k revenue forecast attached. The full analyst loop walked end-to-end: identify, triage, test, decide, ship.

Built as the case study a hiring manager actually wants to read. Not a tool tour, not a dashboard screenshot, not a chart gallery. A real operating problem, four hypotheses tested honestly, a control-group caveat called out explicitly, and a decision with rationale and revisit dates.

Loop Identify → Triage → Test → Decide → Ship Length ~7 min read Method Hypothesis triage · A/B test design · forecast
Walk the decision → See the tool →
Q3 2025 · 30-day controlled test
5pp
Utilization gap that triggered the test. One vehicle off the optimal Q3 calendar by 5 percentage points cost an estimated $14k in unrealized revenue.
identify · triage · test · decide · ship
Live · Operating

Vantage Fleet OS

A single-page analytics dashboard built for a Turo rental business of any size. Drop in any host's CSV export and it answers five real operating questions plus surfaces a triage list of issues worth investigating, with recommended actions.

Built because the Turo native dashboard answers what happened but never why. This one runs Welch's t-tests, OLS regression with R², and z-score outlier flags entirely client-side. 1,700 lines of HTML, no framework, no backend.

Stack Vanilla JS · Plotly · PapaParse Stats Welch's t · OLS · z-score · 95% CI Privacy 100% client-side
Open dashboard → Read the case study →
Published · Long read

Two cheap cars beat my Tesla.

A 12-month investigation of my own fleet. The two cheapest vehicles on the lot, a 2014 Volt and a 2018 Yaris, returned 191% and 83% of capital while the Tesla Model 3 Performance returned 28%. The story walks the data trail that got me there and the operating decision it forced.

Written as a four-chart narrative, not a slide deck. Each chart has a finding-led caption, a method note, and a designed-in answer to the obvious counter-arguments. Same dataset as the dashboard, different job.

Format Narrative analysis · 4 charts Length ~6 min read Method Yield, ROI, utilization, opportunity cost
Read the data story → See the live dashboard →
12-month investigation · same fleet
191%
Return on capital from the cheapest car on the lot. Volt: 191%. Yaris: 83%. Tesla Model 3 Performance: 28%. Same period, same operator, very different math.
4-chart narrative · ~6 min read
Live · LLM agent demo

Fleet OS Agent

A natural-language interface over the same fleet dataset. Ask "which car earned the least per booked day in Q4?" and the agent picks the right tool, runs the query, and answers with both the chart and the SQL it generated.

Built to test how far a small in-browser agent can get with good schema priming and a tight set of typed tools. Same data model as the dashboard, very different question-answering surface. A tiny preview of where I'd take analyst tooling next.

Stack Vanilla JS · LLM tool-use · Plotly Pattern Schema-primed agent · 6 typed tools Surface NL → SQL → chart
Try the agent → See the dashboard →

Also in the lab

02 / EXPERIMENTS

Smaller experiments and works in progress. Some of these are answer-shaped questions; others are tools I built for myself first to see if they were worth polishing for anyone else.

Open notebook · Options research

IV Crush, Quantified

A working paper backtesting the implied-volatility crush trade across the S&P 100 over five earnings cycles. Tests the conventional wisdom: does the post-earnings drop in implied vol actually pay, or does the move in the underlying eat the edge? Real options chain data, vectorized in Python, transaction costs separated out so the headline P&L isn't lying.

Read the notebook → Build journal · Trading systems

MES futures automation

A Python bot that places defined-risk MES futures trades through the Tradovate API based on a small set of intraday rules. Currently runs with my own capital. Build journal covering architecture, risk model, and the operational lessons I've learned running it live.

Read the journal → In-build brief · Product

DayFlow

A spending tracker built around one question: where did the money actually go today? React + Supabase, PWA, manual-entry-first, no bank linking in v1 on purpose. Schema, wireframe, and the case for shipping the small thing first.

Read the brief →
BO
Brad O'Haire Mountain House · CA Est. 1999
Based South San Jose, CA
Today Tesla, Optimus program
Building Vantage Ventures (holdco)
Studying Data Analytics, SNHU + CISA
Looking for Fleet Ops · Operations Analyst · TPM

I run operations on a 24/7 robotics pipeline at night, and a five-vehicle commercial fleet on the side.

At Tesla I'm the Data Operations Lead on the Optimus program's night shift. I keep the data collection pipeline running across a multi-cell floor, automate the redundant pieces out of my team's day, and build the dashboards we use to spot drift before it becomes a stoppage. It is operations work for an autonomous-robotics program, and it is portable to operations work on any autonomous program.

Outside Tesla I run a five-vehicle commercial fleet on Turo. Twenty-six months in, approaching All-Star Host status, with a real P&L that has rewarded me for being right about utilization, pricing, and capital reallocation, and punished me for being wrong. I built the analytics stack myself because nothing off the shelf answered the questions I actually had.

I learn fastest with skin in the game. I'm looking for the role where the operational analysis I produce changes what the team ships next, not just what shows up in the next slide.

Trajectory: fleet operations or operations analyst in the next 12 months, technical program manager on a 2-3 year arc. Strongest fit at autonomous-systems and logistics companies where someone needs to keep a real-world operation running and tell the rest of the org what's going on inside it.

Things I'm doing right now besides applying for jobs: finishing a CPTC cert, two SNHU classes toward the Data Analytics BS, shipping DayFlow, and writing up the next case study from the fleet.

How I work

03 / TOOLING

Analysis

  • Python pandas, polars
  • SQL postgres, bigquery
  • R occasional
  • Jupyter, Quarto
  • Matplotlib, Plotly, Seaborn

Build

  • JavaScript / TypeScript
  • React, Next.js
  • Node.js, Supabase
  • Tailwind, vanilla CSS
  • Vercel, Cloudflare

Operate

  • Tessie API fleet telemetry
  • Tradovate, Robinhood, Tastytrade APIs
  • Notion, Obsidian, Airtable
  • Linear, Asana
  • Looker Studio, Metabase

Background

  • CISA Associate 2025
  • Harvard CS50P
  • SNHU BS Data Analytics in progress
  • CPTC in study
  • Google Tech Writing

Let's talk shop.

I read every email. If you're hiring for fleet operations, operations analyst, or TPM roles, especially at an autonomous-systems or logistics company, the fastest way to a real conversation is a short note about the team and the problem.

Email bradtohaire@gmail.com Resume request via email → LinkedIn linkedin.com/in/brad-o-haire ↗ GitHub github.com/Veeiy ↗