[Finterest] Meet the young Filipina teaching AI to read crumpled payslips for loans
MANILA, Philippines – A loan application in the Philippines doesn’t always come with crisp bank statements and complete payslips.
It may instead be a blurry rural bank form, a handwritten record folded one too many times, or screenshots of GCash transactions from someone outside the formal banking system. But somewhere in that jumble could be evidence that the borrower can repay.
Young Filipino founder Carmel Limcaoco wants AI to do the digging.
Limcaoco and her co-founder — longtime best friend and Stanford classmate Rhea Malhotra — founded Kita, an artificial intelligence platform that helps banks, fintech companies, microfinance institutions, and other lenders process loan applications.
Kita is not a lender. Working behind the scenes, its “AI credit officer” requests documents based on a borrower’s circumstances, spots missing information, and follows up through Viber, WhatsApp, or SMS.
“Lenders in the Philippines face two problems at once. They are badly backlogged, and they lack sufficient data to decide with confidence. We address both,” Limcaoco told Rappler in an interview.
Kita said its technology can read more than 50 types of documents, from PDFs and e-wallet screenshots to handwritten notes and photos of crumpled paper. Using vision-language models, it can understand Tagalog, Taglish, and Cebuano, alongside Spanish, Bahasa, and other languages.
“We read documents the way a person does, from what is on the page, so a rural bank’s handwritten passbook or a format nobody has encoded works on first submission,” Limcaoco said.
Kita extracts the figures, flags inconsistencies, applies the lender’s credit rules, and drafts a credit memo for human review. Each figure is linked back to its source document.
In its first five months, Kita said it processed more than $130 million in loan volume, over 60% of it from the Philippines.
One client is TRBank, a rural bank whose four years of lending history sat in thousands of paper folders. Kita said it converted the archive into searchable data in under three days, with accuracy verified at above 97%.
![[Finterest] How your e-wallet history could help you get a cheaper loan](https://www.rappler.com/tachyon/2026/05/DIGITAL-TOWN-SQUARE-graphics.jpg?fit=449%2C449)
From a one-bedroom apartment to Y Combinator
Kita began with a friendship, a shared one-bedroom apartment in Boston, and eventually, a trip to Manila.
Limcaoco and Malhotra met in 2020, when the pandemic threw them into unusually close quarters. They later studied computer science at Stanford and began searching for something they could build together. In Manila, they found it in an unlikely place: loan paperwork.
Raised in Manila, Limcaoco received United Nations recognition at 16 for her work with One Million Lights Philippines. She later studied at Stanford, worked on audio and music products at Apple, and helped launch Kaya Founders’ fellowship for aspiring product managers.
Malhotra entered research at 13, contributed to Pfizer’s COVID-19 vaccine effort at 17, and studied computer vision and robotics at Stanford. She deferred a Princeton doctorate in AI to build Kita.
That pedigree has attracted some heavyweight believers. Kita raised $4.5 million in a seed round led by BoxGroup, with Lisa Gokongwei-Cheng joining as an angel investor alongside Kaya Founders, the venture capital firm she co-founded and whose funds her family helped anchor. Other backers include Apex Star Capital, the family office of Xiaomi co-founder Lin Bin.
Kita also passed through Y Combinator, the start-up kingmaker that backed Reddit, Airbnb, and Stripe at their earliest stages. The company says Limcaoco is the first Filipina founder admitted to YC in over five years.
How far will Kita go?
But competition is already brewing, sometimes under the same investor’s roof. Kaya Founders also backs Moneta, an AI platform that helps banks analyze corporate borrowers’ documents and produce lender-ready credit assessments. Not exactly the same but pretty close to Kita’s turf.
Limcaoco isn’t too bothered.
“We would expect any fund serious about fintech infrastructure to look at more than one company in a category this size,” she said.
She argues that Kita stands apart by reading unfamiliar formats without advance configuration, tracing every output to its source, and choosing “the hardest path first” by working with traditional banks and large microlenders carrying decades of legacy records.
Kita will remain close to San Francisco’s AI talent pool while hiring in the Philippines, where it says it is investing most deliberately. Its new funding will expand its engineering team and strengthen its underwriting and fraud-detection tools as it pursues customers across Southeast Asia, Latin America, the US, and Africa.
“A rural bank in the Philippines and a microlender in Mexico are not the same institution, but the sequence they move through is similar enough that lessons carry,” Limcaoco said.
Even with those global ambitions, the founders draw a line around what the technology should do. Yes, AI can read the rumpled notes, find the missing figures, and write a digestible credit memo. Tedious work, but largely routine.
What AI should not have is the final word on a decision that could shape someone’s financial future.
“Decisions that deeply affect people’s livelihoods should always involve human oversight,” Malhotra said. – Rappler.com
Finterest is Rappler’s series that demystifies the world of money and gives practical advice on managing your personal finances.
