Aman Singh

Aman Singh leads AI at Arca, a New York financial services firm that combines human advisors with software agents. His work focuses on agents that coordinate client requests over several days, preserving context through pauses and involving people when their judgment or action is needed.

At Stripe, Singh worked on consumer payments and fraud prevention. He was part of the team that launched Link, Stripe’s first consumer product, which speeds up checkout by reusing saved payment details. He also worked on Radar, which helps businesses detect and block fraudulent payments.

He later worked in applied AI at Stripe, helping establish its company-wide framework for AI agents. He shipped agents that researched the web, operated browsers through checkout flows, and automatically assembled evidence for payment disputes.

At Arca, Singh and his colleagues work directly with the advisors who use their software. The firm provides financial services itself, giving engineers direct feedback from the people doing the work. The team’s platform includes sandboxes where agents run financial calculations and interfaces generated for a user’s particular context.

Singh’s approach to long-horizon work centers on the pauses between actions. A client may take time to reply, or a transfer may need days to settle. Arca’s platform lets agents respond to incoming emails, calls, documents, and market changes. An agent can ask to be notified when a client replies and schedule a later check in case the reply never comes. Persistent memory keeps the client’s history and unfinished obligations available when the agent resumes.

Singh distinguishes work an agent must remember from work a person can do now. Future obligations remain in memory; an actionable step becomes a task for the appropriate person. The team’s shared workspaces let colleagues discuss a plan with an agent and complete the steps that require them, such as approving the plan or an email to a client.

Singh and his colleagues also standardize the firm’s workflows and give agents those operating procedures. Colleagues’ explanations of why an action was right or wrong become material for evaluating and improving the agents. His ambition is to steer the firm by changing its agents’ goals. A goal of proactive tax planning, for example, could have agents prepare analyses and client emails for colleagues to approve before a client has to ask for help.