About Fresco
America is in a housing crisis that starts with one simple problem: It’s almost impossible to predict how much buildings will cost.
Fresco is solving this problem after we (Akhil and Arvind) saw estimators lose $MMs in bad bids. We’ve built the category leader in estimating software, scaled past $1M in ARR in just 5 months, and work with some of the largest players in the industry.
The Technical Problem
Construction blueprints aren't clean data. They're enormous, spatially complex PDFs full of ambiguous annotations, inconsistent scales, and domain-specific shorthand that no frontier model handles well.
You'll be building systems that
- Parse and extract structured line items from GBs of messy blueprint PDFs at scale
- Help LLMs turn unstructured construction documents into accurate estimates
- Close the loop with users: talk to an estimator in the morning, ship a fix by afternoon
Our Process
We obsess over getting the data model right so the rest of the code flows easily. We create pull requests for visibility but skip code review. We optimize for shipping speed, not process.
Who You Are
- You love pushing frontier models to do things they aren’t supposed to
- You're a mid-career engineer who's energized by ambiguity and ownership
- Low ego, high output
- You want meaningful compensation for hard work
Why This Role
- Akhil trained some of the earliest LLMs at MIT and various startups— you’ll sit a desk over
- Small team at an inflection point: your code ships to real users immediately
- Backed by Y Combinator, SignalFire, and Bessemer Venture Partners, who have backed Anthropic, OpenAI, Waymo, etc.
Schedule & Location
10am to 7pm, Monday through Friday, in our SF office. Many of us are here Sundays too. This is a full-time, in-person role.
Job description provided by Fresco.
View the original posting
