Inspiration

LookThrough grew out of Mike Koch’s conversations with fund teams while building Osmy, an AI-native fund administrator for private-market funds. Investors struggled to turn scattered reports into a clear picture of their holdings. Mike brought 11 years of product experience, including work at Amazon and Guidepoint; Sarah Yu, founder and AI engineer, brought experience building conversational AI and user-facing applications.

What it does

LookThrough gives limited partners a personal workspace for understanding their investments. Investors set their priorities, ask what changed, explore future cash needs and open the evidence behind reported figures. They can compare corrected documents, review questions drafted for their GPs, and use akta to explore public company information alongside their fund reports.

How we built it

We built the interface with JavaScript, HTML, CSS and interactive SVG charts, supported by a Python backend. A structured evidence pack connects portfolio figures to documents, calculations and ledger entries. We use pdfplumber for supported PDF extraction, Nicrron for configurable AI responses, and akta for company search, enrichment and news. The public demo uses synthetic data and scripted responses; the local app supports the backend integrations.

Challenges we ran into

The hardest challenge was keeping the experience simple without hiding important distinctions. A capital call requests money, a bank transaction confirms payment, and a later statement reports the contribution. Corrected documents need their history, forecasts need explicit assumptions, and external company information must stay separate from fund-reported figures.

Accomplishments that we're proud of

We connected portfolio explanations, source evidence, document history and follow-up requests in one workflow. Investors can compare a corrected capital call with its original version, inspect the matching bank payments, and see cash needs update when a simulated $75,000 call arrives. Our akta integration adds public context without overwriting the portfolio’s reported numbers.

What we learned

Transparency requires more than access to documents. Investors need to understand what a number represents, when it was reported, how it was calculated and what remains unanswered. Designing around those questions helped us make the assistant easier to understand and its answers easier to inspect.

What's next for LookThrought

We want to test the experience with LPs and family offices, expand document extraction, and add persistent history and permissioned connections to email and fund administrators. Our goal is a personal assistant that remembers an investor’s priorities and explains relevant changes as new information arrives.