Before ChatGPT became a household name, a quiet public bet on small businesses and AI
changed the trajectory of commercial real estate.
Long before generative AI dominated headlines, the Commonwealth of Massachusetts
implemented a consequential initiative. Rather than waiting for large corporations to
bring AI innovation to the state, Massachusetts decided to push that power directly to
small and mid-sized businesses, those that likely couldn't afford a research division or a
team of data scientists, but had smart ideas and problems technology could solve.
The vehicle was the AI Jumpstart Program, launched in April 2021 by the Innovation
Institute at the Massachusetts Technology Collaborative (MassTech). The structure was
ambitious: the Commonwealth awarded Northeastern University a $2.2 million grant,
which Northeastern matched with $2 million of its own — creating a $4.2 million public-
private fund dedicated to connecting Massachusetts companies with world-class
academic AI expertise and computing infrastructure they never could have accessed
otherwise.
Companies that joined the program gained access to state-of-the-art multi-architecture
computing clusters, expert faculty from Northeastern, Tufts, and Boston University,
hands-on AI workshops, and Northeastern's $25 million Colosseum facility, which
contains the world's largest wireless emulator.
One of the 18 companies was a Boston startup still in its first year of business:
PredictAP.
Founded in 2020 by David Stifter and Russell Franks, backed by a founding team with
engineering backgrounds at Apple, Google, HubSpot, and Datadog, PredictAP was built
to solve one of the most persistent pain points in commercial real estate operations: the
manual, error-prone process of coding invoices across complex multi-entity property
structures.
For anyone unfamiliar with how real estate AP works: every invoice that arrives at a property management firm needs to be coded and assigned to the right property, cost
center, owner, and general ledger account. Multiply that across hundreds of properties,
dozens of vendors, and thousands of invoices a month, and you have a task that is
simultaneously critical, tedious, and highly error-prone when done manually.
Traditionally Optical Character Recognition (OCR) and indexing services could digitize invoices, but couldn't code them. PredictAP was created to change that process and make it faster and more intuitive than ever before, thanks to the power of AI.
The challenge was significant. Invoice coding isn't a one-size-fits-all problem — each
property, owner, and vendor can introduce different rules and patterns. Building AI that
could learn continuously, adapt to each client's unique logic, and operate reliably in live
production environments required more than a talented engineering team. It required
rigorous scientific thinking.
Through the AI Jumpstart Program, PredictAP was matched with Dr. Mirek Riedewald,
a professor of computer science at Northeastern University, and an expert in databases,
scalable analytics, and query visualization. The match wasn't automatic. As PredictAP
CEO David Stifter recalled, when they first pitched their idea to the program, "most
professors liked the larger ideas of using AI to cure cancer or solve global warming, so it
was crickets (for us)."
Dr. Riedewald heard something different. He recognized the unique business case
depth in PredictAP's complex industry-specific challenge: multi-entity property
structures, property-specific invoicing rules, historical payment patterns, and the need
for AI models that could learn and adapt continuously. This was exactly the kind of
applied, human-scale problem that theory alone couldn't solve, and precisely what his
students were going to face when they hit the job market after graduation.
Over several months of collaboration, the engagement became as much an education
for Dr. Riedewald as it was for PredictAP. He saw, up close, the gap between what can
be proven in a lab and what can be deployed effectively in a platform used daily by real
estate finance teams across the country.
What made the PredictAP/Northeastern collaboration more than a community
connection was the input quality each side brought to the table, and the discipline that
let each partner lead where they were strongest.
Dr. Riedewald contributed the scientific process: structured research methodology,
rigorous model evaluation, and an academic's instinct to interrogate assumptions before
accepting results. He brought the intellectual firepower to take PredictAP's AI
development beyond what a startup engineering team might pursue on its own.
PredictAP's engineering team, led by Chris Antenesse (formerly of Apple), brought
something equally essential: deep expertise in building software that works at scale in
the real world, He pushed the product to be architected for reliability, designed for
continuous learning, and deployed successfully in live production environments.
The combination produced results that neither party could have achieved alone. Today,
PredictAP's platform:
• Processes nearly 3 million invoices annually
• Updates its AI models daily to capture evolving coding patterns and client-specific
rules
• Delivers a documented 4–5x ROI in time savings for customers
• Integrates seamlessly with Yardi Payscan, Nexus Systems, and other leading AP
platforms
• Serves 80+ customers, including Bridge Investment Group, The RMR Group,
Starwood, and CA Ventures
On the business side, the academic partnership helped fuel a trajectory that speaks for
itself: an $8 million Series A in November 2023, followed by an additional $5 million
round led by RET Ventures, and over 100% year-over-year revenue growth following
the Series A.
PredictAP's journey through the AI Jumpstart Program is more than a startup success
story. It is a proof of concept for a model of innovation that Massachusetts has
continued to scale.
In 2024, the Healey-Driscoll Administration launched the Massachusetts AI Hub — a
statewide initiative carrying forward the same public-private philosophy of AI Jumpstart,
now at a far larger scale. The Hub has anchored a $31 million state investment (the first
phase of a planned $120 million commitment) to build the Artificial Intelligence Compute
Resources (AICR) cluster at the Massachusetts Green High Performance Computing
Center in Holyoke and has brokered a partnership with Google to offer free AI training to
every Massachusetts resident.
The throughline from AI Jumpstart to the MA AI Hub is a commitment to the partnership
between government, universities, and entrepreneurs who are willing to bet that a hard,
specific, real-world problem is worth solving.
PredictAP's story is a compelling example of what becomes possible when public
investment, academic expertise, and entrepreneurial vision align. The Massachusetts AI
Jumpstart Program gave PredictAP the academic partnership and computational
resources it needed to accelerate its AI development at a critical early stage, well before
generative AI became a household term.