Akshay Petta Left BlackRock to Build the Systems Behind Reliable Financial AI

After using code to transform investment analysis, the former strategist moved into engineering and began building regulated financial platforms from the ground up.

A global market dashboard once took approximately six hours to complete inside Akshay Petta’s division at BlackRock. Petta rebuilt the process in Python and reduced the workload to about 30 minutes. The project gave him his first practical example of how code could reduce repetitive financial work while leaving the decisions themselves with the people who understood the financial context.

Petta had joined BlackRock after receiving a full-tuition scholarship to the London School of Economics. At 22, he was part of a three-person investment strategy team supporting a $6 billion multi-asset fund. He advised major insurers and built Python tools used to model more than $100 billion in assets for some of the firm’s largest institutional clients. His route there had not been straightforward: Petta was born in India and moved to the UK at age 12, where he attended state schools before earning his scholarship to LSE.

“Watching highly trained people spend so much time on mechanical processes made me curious about what could be rebuilt,” Petta said. “The dashboard project showed me that code could remove hours of work while leaving the important decisions with the people who understood the financial context.” He taught himself to code while working as an investment strategist, initially as a practical way to improve internal processes. The tools he created eventually became a larger part of his work and showed him how closely technical choices could affect the professionals relying on the results.

“If you are in finance and curious about engineering, your domain knowledge counts for more than you may think,” Petta said. “The code can be learned.” Over time, he became more interested in building systems than using their outputs. Leaving BlackRock meant walking away from an established finance career and entering the startup world as a self-taught engineer who still had to prove his technical ability through what he delivered.

In May 2024, Petta became the founding engineer at Wyzr, a UK financial technology company developing an AI CFO platform for small businesses. He built most of its data and AI infrastructure, including the minimum viable product that underpinned a £400,000 funding raise. His responsibilities included creating an FCA-regulated Open Banking pipeline handling customer financial records, which meant reliability had to be designed into the architecture from the beginning.

“At Wyzr, I was responsible for systems where a bug could corrupt a customer’s financial data,” he said. “There was nobody above me to hand the problem to. I had to build carefully, test the work, and understand what was happening across the platform.” Petta carried systems from their technical foundations into production, working across data, artificial intelligence, and product requirements. The experience reinforced his view that documentation, testing, and clear interfaces can support speed rather than stand in its way.

In January 2026, Petta joined Bead AI as its first hire and founding engineer. The a16z-backed startup builds AI agents that automate Sarbanes-Oxley audit testing for publicly listed companies in the United States. At Bead, financial documents similar to those Petta once worked with became the material his software had to interpret, often across spreadsheets containing complicated formulas, external references, and relationships spanning hundreds of files.

Petta designed and built the spreadsheet-agent infrastructure that allows Bead’s agents to process unusually large and interconnected evidence sets. His work includes the caching, memory-management, and distributed-processing layers needed to handle that material reliably before producing working papers that human auditors can review. “Financial spreadsheets are rarely simple grids of numbers,” Petta said. “Their meaning can sit in formulas, formatting, file relationships, or context that is not obvious from one cell. An agent has to handle that complexity and still show how it reached a conclusion.”

Bead’s agents have reduced some SOX control-testing processes from days to minutes, and Petta’s spreadsheet infrastructure was central to winning the company’s first client, described as one of the largest publicly listed companies in the world. His earlier career still influences how he approaches the work because he spent years using financial models before building agents that must interpret financial evidence. He understands why an auditor needs traceability and why systems operating in regulated environments must allow professionals to examine how a conclusion was reached.

Outside his operating roles, Petta has also been selected for external evaluation work. Innovate UK named him an Expert Assessor in its AI, Digital and Advanced Computing category. Over this summer, he completed six peer reviews of AI and machine-learning research for IEEE Access; three of the invitations came from Associate Editors, while the other three were signed by Editor-in-Chief Prof. Mehrdad Saif. IEEE Access also issued him a Certificate of Reviewer Recognition signed by its Managing Editor. In August 2026, MassChallenge appointed him to its expert community as a judge and mentor following an application and screening interview.

Petta’s goal remains reducing the mechanical evidence checking that occupies audit, compliance, and finance teams while preserving the judgment professionals bring to consequential decisions. “At BlackRock, I learned how financial professionals work with information and why trust is essential,” he said. “Engineering gave me a way to redesign the processes around that work. I am still solving financial problems, but now I am building the systems that help other people solve them.”

The six-hour dashboard was Petta’s first example of what that approach could accomplish. His work has since expanded from one internal process to infrastructure designed for enterprise audit systems. The scale has changed, but the principle remains familiar: understand the work first, then build a dependable way to perform the parts that no longer need to consume a person’s day.