iPulse AI

I was 15 when I bought my first stock: Apple, in 2007.

I did not have a sophisticated model or an investing framework. I was simply fascinated by the idea that I could study a company, form a view of its future, and own a tiny piece of that future.

That first investment became a much larger responsibility. Over time, I began helping manage my family's global wealth portfolio across markets and asset classes. Investing stopped feeling like a game. These were not abstract tickers on a screen. They represented years of work, security, and choices my family might have later. Every decision felt heavier.

The research was endless. I would move between annual reports, valuation models, macroeconomic data, industry developments, geopolitical risks, and competing scenarios. One answer created five more questions. Even after hours of work, I knew there were important angles I had not considered.

At the same time, I was specializing in artificial intelligence at university. I later worked at JPMorgan and consulted on AI topics at PwC and Accenture. I saw the enormous potential of AI, but I also saw the distance between an impressive demonstration and a system you could trust around serious decisions.

For years, AI was not ready. It could automate narrow tasks, classify information, and find patterns, but it could not reason through an investment thesis with enough depth or flexibility.

Then GPT-3 arrived.

It was far from perfect, but the direction became clear to me. The foundations were finally there. For the first time, I could imagine machines reading and comparing more information than any individual analyst could hold, approaching the same asset from several investment philosophies, and continuously stress-testing one another's conclusions.

I became convinced that AI would eventually reason better than any single financial analyst—not because it would become infallible, but because it could combine breadth, speed, memory, and many independent perspectives at a scale no person could reproduce.

I searched for the platform I wanted to use. What I found were polished answers, unexplained scores, and forecasts without an inspectable history. The workflow behind the conclusion was usually hidden. When the models or methods changed, the past often disappeared with them.

That was the part I could not accept.

A forecast without a timestamp, methodology, evidence trail, and historical record can always be rewritten after the fact. It may look intelligent, but it cannot earn trust.

So I started building iPulse AI.

iPulse AI did not begin on its current architecture. We initially built on AWS, then moved to GCP as the research workload expanded across market-data pipelines, parallel AI-agent runs, forecast history, and the public application. For our workload, combining Cloud Run, BigQuery, Firebase, and the wider data platform reduced operational friction and made experimentation easier. Being accepted into the 2024 Google for Startups AI Startup Program—and receiving $2,000 in credits—gave a self-funded product valuable room to test ideas before knowing which architecture would survive.

The goal is not to replace human judgment or pretend that AI knows the future. It is to make investment research broader, faster, and much more transparent. Independent AI agents examine an asset through different lenses, expose their reasoning, disagree with one another, identify risks and drivers, and produce forecast paths that can be reviewed later rather than quietly forgotten.

The consensus matters, but the disagreement matters too. The strongest idea is not always the one with the loudest bullish score. Sometimes it is the one that survives the most serious stress tests.

For us, “open” means making the research process inspectable: the methodology, architecture, configurations, evidence, historical forecasts, evaluations, limitations, and lessons—including what did not work. It does not mean pretending our models are certain, or hiding failures until the next version looks better.

iPulse AI is the platform I wish I had when I bought Apple at 15. More importantly, it is the platform I need now, when investment decisions carry real responsibility.

I want investors to be able to ask more than, “What does the AI think?”

Why does it think that? Which assumptions matter? Where do the agents disagree? What would break the thesis? And when we return to this forecast in a year, will the original reasoning still be there?

That is the standard I am trying to build toward.

iPulse AI is now live at https://ipulseai.com. If you manage your own capital, your family's wealth, or research for clients, I would genuinely value your most critical feedback: what evidence would you need to see before allowing AI-generated research to influence a real investment decision?

This is research and decision-support software, not financial advice. The models can be wrong. The point is to make their reasoning—and their mistakes—visible.