Case study · a 16-month contract
The data behind an AI forecasting platform
For 16 months I was a contract AI engineer at Vico Technologies, an enterprise AI-forecasting platform. I worked on the forecaster and on everything that feeds it: collecting the news, translating it and making it searchable.
The job
A forecaster is only as good as what it reads. This one reads the news. Thousands of new articles a day, many of them not in English, have to be collected, translated, stored and found again quickly.
What I built
- Collection. Daily scrapers and ingestion pipelines across hundreds of sources.
- Search. I prepared a corpus of more than a million news articles and tuned the database behind it so retrieval stays fast at that size.
- A corpus for one customer. For one end client, a Global 100 law firm, I built a retrieval configuration around economic, political and government sources.
- Translation. Hundreds of thousands of foreign-language articles translated with AI. A human translator independently checked the Chinese output and found it more than 95% accurate.
- Analysis. A layer that classifies article text for sentiment, bias and political leaning against defined criteria.
- The forecaster. I implemented, researched, improved and continuously tested the AI forecaster itself.
Who used it
The platform's customers included a global financial-data and media company, a Global 100 law firm and a federally funded research organization. They were the platform's customers, not mine.
Why this matters for your business
The same skills apply at a small scale: collecting information from many places on a schedule, making your own documents searchable, and having a person check the output before anyone relies on it. I ran smaller client contracts alongside this one the whole time.