Klimacap · AI Developer Intern
I built an AI pipeline that turns an M&A analyst's short request into a finished company or industry report in Word, with every claim cited. It cut report writing from hours per company to minutes.
01The problem
M&A analysts spent hours on each company pulling scattered research into a readable brief. Their judgment wasn't the slow part. Gathering, structuring, formatting and citing everything the same way each time was.
02What I built
I was the only engineer at the firm. I built a tool that takes a short request from an analyst and runs it through stages: company profile, industry context, competitors, then a summary. Each stage feeds the next automatically, so nobody has to type the follow-up prompts.
03The output
Reports come out as formatted Word documents, the format analysts actually send on, with citations linked to sources and a bibliography at the end. There's no copying from a chat window into a template.
04A decision that mattered
Every claim in a report links back to its source. An M&A report nobody can verify isn't usable, so I built the citations into the pipeline instead of adding them at the end.
05How it links to my other work
This is where the idea behind GUS started: generate from sources, and check as you go.