AngelAi is an artificial intelligence platform developed by Celligence to execute complex financial transactions rather than simply answer questions or generate content. The platform combines a deterministic Transactional Language Model, specialized Angel Agents, and Warranted Intelligence to collect and reconcile information, apply rules, make decisions, and carry transactions through to completion. AngelAi has been operating within Sun West Mortgage since 2019, giving its technology a production environment for handling real financial transactions. Pulse 2.0 interviewed Sun West Mortgage Company CEO and Founder/CEO of Celligence and AngelAi, Pavan Agarwal, to learn more about AngelAi’s development, approach to deterministic AI, and ambitions beyond mortgage lending.
Pavan Agarwal’s Background
When asked about his background, Agarwal shared:
I grew up around the mortgage business. My father, Hari Agarwal, founded Sun West Mortgage in 1980, so from a young age I had a front-row seat to the industry.
But my own interests were always rooted in technology, mathematics, and engineering.
I took over Sun West in 2008, just as the financial crisis was unfolding. Bear Stearns had collapsed, the mortgage industry was under enormous pressure, and I was inheriting a company at probably the most difficult moment imaginable.
It forced me to think differently about how a mortgage company should operate while maintaining our core values of fair and affordable housing.
These core values are why we, unlike most of our peers of that time, thrived during the subprime crisis.
That led me to build Celligence, the financial technology company behind AngelAi, with Sun West becoming the proving ground where we could develop the technology against real financial transactions.
Today, I serve as CEO of Sun West and founder and CEO of Celligence and AngelAi, but I still think of myself primarily as a technologist and a builder.
How AngelAi Started
When asked how the idea for AngelAi came together, Agarwal explained:
AngelAi grew out of a problem I was trying to solve inside Sun West.
I started looking at every part of the mortgage process and asking why so much work still required people to manually collect documents, interpret information, move data between systems and make repetitive decisions.
We began automating those individual tasks, but I eventually realized that automating one step at a time wasn’t enough.
A mortgage is an interconnected transaction, and the technology needed to account for the entire process.
That became the foundation for AngelAi.
It has been operating inside Sun West since 2019.
What started as an effort to make our own company operate better ultimately became a platform for using AI to execute complex financial transactions at scale.
Favorite Memory
When asked about his favorite memory working with the technology, Agarwal said:
My father Hari never celebrated or expressed emotions of any kind.
In 1982, I watched the slight crack in his stoic presence as the corners of his mouth lifted and a smile momentarily glanced across his face when my PC software printed the first set of Truth-in-Lending disclosures and they were handed to a consumer on the spot.
In retrospect, I now understand why his emotions were contained because he also was fully aware of the difficult road ahead of us as the first lender to deliver loan origination on a PC.
We needed to:
- Convince Freddie, Fannie, Ginnie and warehouse bankers to accept the loans produced by this software.
- Continue the investment into further development.
- Stay the course through the waves of detractors.
Core Products And Features
When asked about AngelAi’s core products and features, Agarwal explained:
AngelAi is designed to do more than answer questions or generate content.
It can understand a financial transaction, collect and reconcile information, apply complex rules, make decisions and help carry that transaction through to completion.
At the center is our Transactional Language Model, or TLM.
Unlike a probabilistic large language model, or LLM, the TLM is deterministic.
Given the same facts and rules, it produces the same result.
That makes decisions explainable, auditable and traceable, which is essential when AI is being trusted with something as consequential as a mortgage.
We have also built specialized Angel Agents that perform functions across the transaction, from processing and underwriting to closing and servicing, within the same system.
Another important feature is Warranted Intelligence.
When AngelAi provides warranted information, Celligence stands behind its accuracy.
We believe accountability has to be part of the product when AI is being trusted with these big decisions.
How AngelAi’s Technology Has Evolved
When asked how the technology has evolved, Agarwal explained:
The process of originating a mortgage is rows of humans making tens of thousands of decisions accurately and rapidly.
These decisions need to be made with uncertain and volatile data.
This is the reason that financial services are slow, error prone, discriminatory, and expensive.
Twenty years ago we imagined an architecture where these decisions are made by autonomous intelligent AI “cells,” also known as Agents, and, circa 2012, we managed to deliver early AI cells into production.
After that, we generated millions of cells and built a conversational layer so people could interact with the technology in natural language.
But unlike a chatbot that is cute but lacks deep intelligence and authority, the AngelAi conversation is connected to the underlying transactional AI.
AngelAi can go from answering a question to doing the work.
More recently, we’ve expanded that model through specialized Angel Agents and begun taking the technology beyond Sun West and its partners, including brokers, realtors, and banks, so anyone can have their own personal AI assistant quickly and easily.
Key AngelAi Milestones
When asked about AngelAi’s most significant milestones, Agarwal said:
AngelAi has grown to more than 415,000 registered users and has supported more than $40 billion in transactions.
More than 5,000 mortgage brokers have signed technology agreements, and 80,000+ households use the platform to make their mortgage payments.
Another milestone has been proving the economics.
At Sun West, we’ve reduced the cost to manufacture a mortgage to under $125, compared with roughly $5,000 across the industry.
To me, that demonstrates that AI can change the cost structure of financial services.
It’s not just about making individual employees a little more productive.
Why Lower Mortgage Origination Costs Matter
When asked why bringing the cost of originating a mortgage down is significant, Agarwal explained:
Because the cost of originating a mortgage affects who lenders can afford to serve.
A borrower with one employer and one W-2 is relatively inexpensive to process.
Someone with multiple jobs, income streams or dozens of documents requires significantly more time and human effort.
That doesn’t necessarily make the second borrower riskier, but it does make the loan more expensive to manufacture.
When you use AI to remove that cost penalty, you can serve more complex borrowers without loosening lending standards.
We recently compared Sun West’s 2025 lending outcomes using AngelAi with national Home Mortgage Disclosure Act data and found that our decline rates were substantially lower than national decline rates across every ethnicity we analyzed.
Complexity shouldn’t prevent qualified borrowers from owning homes.
Based on 2025 data, when AngelAi is applied across the industry, an additional 1.1 million American families would be homeowners.
Customer Success Story
When asked to share a specific customer success story, Agarwal highlighted a traveling nurse:
One that has stayed with me involved a traveling nurse who earned good money but had worked assignments across the country and accumulated a large number of W-2s from different employers.
AngelAi was able to reconcile those income sources, determine her eligibility and help her become a homeowner.
Funding
When asked about funding, Agarwal shared:
I funded the development of this technology largely myself, investing approximately $500 million over the years to build the platform and prove it in production.
More recently, Mortgage Treasury invested $100 million in Celligence and AngelAi.
That is followed by its $1 billion commitment to purchase mortgages originated through AngelAi and Sun West.
We don’t disclose revenue, but I think the transaction volume gives a useful indication of scale.
For me, the important part of the funding story is that we spent years investing in and proving the technology in the real world before bringing in significant outside capital.
Competitive Differentiation
When asked what differentiates AngelAi from its competition, Agarwal explained:
The biggest difference is that AngelAi was built to transact.
It combines deterministic AI, end-to-end execution, and Warranted Intelligence, so the technology can do the work, explain how it reached a decision and stand behind the result.
We’ve also had the advantage of proving it against real financial transactions since 2019.
That production history is difficult to replicate.
Future Goals
When asked about AngelAi’s future goals, Agarwal said:
Our immediate goal is to take what we’ve proven at Sun West and make it available much more broadly to brokers, lenders and other financial institutions.
The recent investment gives us additional resources to do that.
Longer term, our ambition goes beyond mortgages and we are expanding into all aspects of banking, insurance and personal finance.
We built AngelAi to execute complex financial transactions, and many of the problems we’ve solved, including fragmented systems, manual processes and the need for explainable, accountable decision-making, exist throughout financial services.
Our goal is to bring the same model we’ve proven in mortgages to other complex financial services.
