
How the PortexAI Datalab saved this startup over 180 developer hours using our Request for Proposal (RFP) feature.
Intro
You know you need data but you don’t know how to source it or whether it even exists.
That’s the position many AI teams find themselves in today. As models become more accessible and commoditized, the bottleneck is no longer in software or writing code; it’s the availability of unique, high-signal datasets. The real moat in today’s AI ecosystem is data that's original, difficult to source, and highly targeted to specific use cases.
But for many of the AI builders we’ve talked to, sourcing novel data is a challenge. This results in missed opportunities, wasted engineers’ time, and a market that moves slower than it should.
This is why we're excited about Requests for Proposals or RFPs on PortexAI’s datalab. RFPs supercharge AI builders by letting them issue requests for datasets that they are seeking. RFPs function as credible demand signals and take the guesswork out of which datasets are most high value.
To showcase the power of RFPs, we recently worked with an early-stage AI real estate investment platform to help them issue a request for a dataset they critically needed for product development.
Background
Despite being an offshoot to the data-heavy world of finance, the commercial real estate (CRE) industry is characterized by disparate and messy information scattered across both public and private sources. With the help of AI, GothamRE (currently in stealth) aims to redefine how real estate data is used by CRE professionals, starting in the largest urban real estate market in the US, New York City. GothemRE recently joined the PortexAI Datalab to source a dataset for its initial product development.
The Challenge
A crucial piece of data when understanding the NYC real estate market is the composition of rent stabilized versus market rate units. Roughly half of the city's two million apartment units are rent stabilized today; however, the breakdown of market and stabilized units by building is not readily available from structured sources like NYC Open Data. The data is publicly available, but is only officially reported in individual New York City Department of Finance property tax bills once a year (which are hosted as PDF files to make things more complicated). Here is an example Statement of Account (SOA) for a building in NYC with 5 rent stabilized units.
An example NYC Department of Finance Statement of Account (SOA) with Rent Stabilized Unit Count (source)
Issuing an RFP
GothamRE issued an RFP on Portex requesting a dataset of rent stabilized apartment counts for over 70,000 buildings in the city as of June 2025. The RFP included clear instructions for dataset construction and data sources, the desired schema, evaluation criteria, and reward amount to be given to a valid responder.
By including an expiration date, they were also able to signal to the market that they needed this data urgently (while also providing a timeline to would-be responders). Within just a week, a data producer on PortexAI’s Datalab was able to produce a structured dataset including the rent stabilized unit counts for the desired universe of buildings. GothamRE was able to evaluate that this data was exactly what they needed after spot checking a sample, with additional assurance from Portex's independent own check. GothamRE then issued a payment to the seller using USDC, and the team was able to seamlessly download the desired dataset and add it to their growing data lake.
Even though the respondent didn't have a real estate background, they knew how to effectively work with the relevant python libraries to parse the PDFs and build a robust system to collect the data based on the instructions written within the RFP.
GothamRE could have spent valuable engineers' time collecting this data internally. While we expect there to always be a "buy vs. build" debate with datasets, issuing an RFP ultimately made the most sense for the team when taking into account speed and cost. Further, an RFP invited competition through a range of solutions from a global market that the team might not even have thought of.
Conclusion
This case study showcases a number of benefits to issuing RFPs on the Portex Datalab when sourcing novel and unique datasets for AI development.
- Clear Costs: All costs are transparent and predictable for both RFP issuers and respondents.
- Speed: Issuers can signal urgency with expiration dates and larger rewards. RFPs recoup engineers’ time from data wrangling and allow them to focus on more high-impact work.
- Invites Innovation: RFPs tap into a global network of respondents that might think of their own unique solutions to source the dataset in need.
- Reputation: Responders can build a reputation on Portex for sourcing unique data.
- Pseudo-anonymity: Buyers can optionally issue RFPs pseudo-anonymously, and sellers can respond pseudo-anonymously.
It is becoming increasingly clear that data is the last true differentiator in a world where software and intelligence is commoditized. We believe RFPs will serve an important role in helping the next generation of AI builders source novel datasets to push innovation forward while rewarding experts transparently and fairly for the datasets they've built and contributed to.
For current PortexAI Datalab users, check out our guide to issue RFPs and respond to RFPs. To access RFP features, sign up for the PortexAI Datalab here.