a16z · Oct 18, 2026 · 24m · 17m saved
With Mario Martone, Head of Applied Research, EliseAI
EliseAI handles about 5 million phone calls a month for housing and healthcare companies, and Mario Martone, its head of applied research, said in an MTS interview published by a16z that running AI at that volume is a harder problem than the string theory he left.
Academic papers can present 90% accuracy as a result. At EliseAI's volume, a model that is right nine times out of ten fails on millions of calls a year, and each failure is someone trying to rent a home or reach a doctor.
"I did not expect that, but even after coming from string theory, I would say the complexity of the problem we're working on now is probably higher now — at least than it was before." — Mario Martone
Mario spent 15 years studying black holes and string theory. He holds PhDs from the University of Naples and Cornell, held postdoctoral posts at the University of Cincinnati, UT Austin and Stony Brook's Simons Center, and joined King's College London's theoretical physics group in 2022 before moving to industry.
The full interview is covered here so you can skip it. 24 minutes of audio, 7 minutes of reading.
Here are the 7 insights that matter.
Key Takeaways
EliseAI handles about 5 million phone calls a month, so a 90% success rate would mean millions of failed calls a year
Lease-renewal predictions run on an in-house neural network, not an LLM
Tests on 5 or 6 conversations make every method look good; the differences only show up across millions
New models are trusted only after they pass EliseAI's own benchmarks, whatever the posts on X say
About 8 months of work on open-source models has put EliseAI's own trained models into production
Leasing can run from a website visit to move-in with no human involved
A maintenance repair still needs a person in the home, a gap Mario thinks could close in a year or two
1. Why He Left Physics
Mario said he could have joined a frontier AI lab, where the research looks much more like theoretical physics. He chose a company whose models act on real decisions.
Mario Martone turned down the research-lab path to work on applied AI
"One other alternative I could have done is, for example, try to join one of these frontier labs... but one thing I really wanted is to have an impact in the real world." — Mario Martone
At EliseAI, the output of a model reaches tenants and patients, not a paper or a benchmark table.
Mario Martone says EliseAI's models make decisions that affect people's lives
"The decisions that the LLMs make for Elise, for our products — they're not just some random things that appear on the screen, but drive impacts on people's life." — Mario Martone
He added that AI changes almost daily, and that the field rewards the same independent, first-principles thinking physics trained him for.
