From the reliability of banking systems to the uncertainty of markets and the possibilities of personal AI. A fieldbook for connecting the dots.
Explore the connections 01Different domains. Shared questions.
Banking systems make reliability a practical question. Every boundary, dependency and recovery path deserves attention.
Failure modes matter in every domain: an interrupted workflow, a fragile research assumption, or an AI agent taking the wrong next step.
I’m Sangamesh. I’m drawn to the space between an interesting question and a working system.
I’m an AI-first generalist focused on human-aligned intelligence. My background spans banking workflows at DBS, CAD collaboration with Autodesk at Pramati, and data protection at Mindtree.
I combine technical, managerial and business thinking, and build iteratively. This fieldbook brings the ideas, the trade-offs and the questions still worth investigating into one place.
Professional background on LinkedInTwo ways into each thread.
At DBS Tech India, I led work on transaction workflows covering payments, securities and deposits. Earlier work included retail and corporate digital customer onboarding.
My work combined project leadership with backend engineering. The questions I carry forward: Where are the boundaries? Which dependencies can fail? How do we make recovery and diagnosis easier?
An exploration of quantitative strategies and systematic research. The interesting part is testing an idea carefully enough to discover where it stops holding up.
The research lens: assumptions, data quality, validation and risk. Strategies remain experimental; a backtest is a question to investigate, not evidence of future profitability.
Exploring how AI agents can become useful collaborators in real work. The goal is practical help, with enough visibility to understand the reasoning behind the next step.
The design questions: What context does an agent need? Which tools should it use? Where does human judgment belong? How do we make its actions inspectable and recoverable?
Open questions, not finished essays.
How do we make a system's failure modes easier to see?
RELIABILITYWhat would change my mind about a research result?
RESEARCHWhere should an AI agent pause for human judgment?
AGENCYBuilding, researching, and learning across the same three threads.
Exploring personal AI workflows. Investigating quantitative ideas through QuantSmart. Carrying the lessons of backend engineering into new products.
An evolving fieldbook — ideas here are allowed to change.
For a conversation about engineering, research or something at their intersection.
Connect on LinkedIn