Profile
The Engineer Who Learned to Bet
There's a specific kind of person who ends up scouting for Andreessen Horowitz. They're usually not the ones who studied venture capital. They're the ones who built things first - real things, complicated things - and eventually figured out that the most important work was finding the right people to build the next ones.
Tony Xia is that kind of person. Before he was writing pre-seed checks at a16z speedrun, he was writing code that let machines see: drone cameras at DJI, autonomous vehicle perception at AutoX, and computer vision pipelines at Arraiy, which Matterport eventually acquired. At the University of Toronto, he dug into AI/ML research before any of those roles. His resume reads like a map of the most technically demanding jobs in the industry.
Then he turned around and built a dating app.
"Redefining international dating one lunch at a time."
- Tony Xia, Co-founder & CTO, Crema Social
One Million Lunches
Crema Social is not what most people expect from a former computer vision engineer. It's a video call platform with food delivery baked in - a place where two people in different cities order food and share a meal over live video. Underneath the interface is Tony's AI translation layer, letting people who don't share a language still share a table.
The idea sounds simple. The execution is not. Building real-time video infrastructure, cross-border food delivery integrations, and AI translation into a consumer product is the kind of thing that takes a team with deep engineering DNA. That's what Tony brought to the partnership with Timothy Chen Chen, a serial entrepreneur with two exits and an MIT pedigree, who serves as CEO.
In eight months, Crema Social crossed one million signups. Not registered users sitting idle in a database - signups for a platform that requires you to actually show up, order food, and be present with another human being. That is a number that opens doors in venture capital.
a16z speedrun Cohort 003 - New York
Crema Social
The app that turned "let's grab lunch" into a cross-border dating ritual. Users connect over live video while both order food in their own city. AI translation handles the language gap. Chemistry handles the rest.
3
Verticals: Dating, Social, Marketplace
The a16z Chapter
Getting into a16z speedrun is hard. The program takes startups from 40+ countries and writes checks up to $1 million. Since launching in 2023, it has deployed more than $180 million to fund over 150 startups. Getting accepted into Cohort 003 as a team with a consumer dating product - not an enterprise SaaS or AI infrastructure play - said something about the traction Crema had built.
But what Tony did next is the part that's harder to script. After going through the program as a founder, he joined a16z speedrun as a Scout. This is not the usual trajectory. You don't typically get funded and then immediately start funding others in the same ecosystem. It requires a firm's trust that you can identify talent the same way you've demonstrated it yourself.
As a Scout, Tony writes $10,000 pre-seed checks to founders he believes have what it takes to enter the speedrun program. It's a small amount by venture standards and an enormous amount in terms of what it signals: a warm connection to one of the most competitive accelerators in Silicon Valley, and a fast lane to up to $1 million if the bet pays off.
Tony Xia - Career Arc at a Glance
AutoX (Self-drive)
Perception
The Pattern Underneath
Look across Tony's career and a thread runs through it that has nothing to do with any single industry. DJI needed cameras that could track subjects in motion at 60 frames per second without GPS. AutoX needed perception systems that could read San Francisco intersections in rain. Arraiy, before Matterport absorbed it, was building computer vision infrastructure for entertainment and media.
Each of those problems is fundamentally about making sense of a chaotic visual world in real time. The engineering is brutal. The constraints are unforgiving. And the person who does that work develops an instinct for systems - for what's structurally sound versus what only looks like it works under perfect conditions.
That instinct translates directly to evaluating startups. When Tony looks at a founding team's technical architecture, or their growth metrics, or how they've responded to a product challenge, he's running the same check he'd run on a vision pipeline. Is this holding because it's built right? Or because the conditions have been favorable?
Venture Scouting
Now Backing the Next Wave
As an a16z speedrun Scout, Tony identifies pre-seed founders with the technical depth and founder DNA to compete for the accelerator's $1M investment and $5M+ in platform credits. He moves fast - decisions in 24 hours, checks written before momentum stalls.
$10K Pre-seed Check
Career Timeline
From Pixels to Portfolio
2016-2019
AI/ML Research - University of Toronto
Deep learning and computer vision research at one of the world's top AI programs, building the technical foundation that would underpin everything after.
2019-2020
Computer Vision Engineer - DJI
Built vision systems for the world's leading drone and camera company. Real-time object tracking, motion estimation, and aerial perception at scale.
2020-2021
Vision Team - Arraiy (acquired by Matterport)
Worked on computer vision infrastructure for professional media and entertainment, before the company was acquired by Matterport, the spatial data pioneer.
2021-2022
Computer Vision Engineer - AutoX
Autonomous driving perception: the hardest real-time computer vision problem in the industry. Built systems to read urban environments at speed.
2023
Co-founded Crema Social (Founder/CTO)
Co-founded with Timothy Chen Chen. Built the entire technical stack: real-time video, food delivery integrations, and AI translation across 40+ languages.
2023-2024
a16z speedrun Cohort 003
Crema Social accepted into one of Silicon Valley's most competitive accelerators. Hit 1M+ signups in 8 months during the program.
2024-present
Scout - a16z speedrun (Andreessen Horowitz)
Writes $10K pre-seed checks to promising founders. Fast-tracks the best into a16z speedrun's pipeline. Based in New York, deploying capital across the East Coast tech ecosystem.
The Details
Five Things Worth Knowing
🍜
Crema Social's premise: order food in your city, turn on the camera, and have dinner with a stranger in another country. Tony's AI handles the translation.
🚁
He built vision systems at DJI - the company that makes the cameras everyone from wedding videographers to the US military uses. Not a hobby project.
🏎️
AutoX put self-driving taxis on the streets of China. Tony worked on the perception stack that made those cars read the road.
📍
Tony is based in New York - not Menlo Park, not San Francisco. One of the few a16z speedrun scouts operating full-time in the East Coast startup ecosystem.
⚡
a16z speedrun scouts move fast by design: decisions in 24 hours, $10K checks that signal a direct path to up to $1M in funding from a16z speedrun.
