The Data Conference 2019: an AI transformation keynote in Athens
On 14 March 2019 Tarry Singh took the stage at the Data Conference in Athens, a netweek and Boussias event on the value of the data scientist, in front of a full auditorium.

Boussias Communications and netweek conferences ran the day under a single line, printed across the opening screen: deconstructing the value of the data scientist in the information economy. Yiannis Mouratidis, billed on that slide as a science and technology editor, took the introduction and the chairing. He was at the lectern by twenty past nine. Many in the room wore simultaneous translation headsets, and a countdown clock sat on the front edge of the red stage, facing whoever was speaking.
Tarry Singh spoke to a full house. His opening slide was a plain whoami: chief executive, founder and AI researcher at deepkapha.ai, LinkedIn Top Voices 2018 for data science and analytics, contributing editor at Forbes on AI in business and society. The clock on the stage edge read 40:42.
The keynote deck, filed since under the name Greece-Keynote, makes one argument. A company becomes an AI company on three pillars: AI maturity, AI skills, and projects picked for return on investment. Two client stories carried the weight. One was an eight-month telco programme that produced a training path for in-house staff and business plans for some twenty AI solutions, worth 2.5 billion euros on the deck's own figures, plus a first AI value map by business domain. The other was a four-month medical pilot, a deep learning cytology classifier separating malignant from benign breast images, at roughly 82 per cent accuracy on a very small dataset.
The rest of the programme stayed technical. One speaker took the room through supervised and unsupervised learning on an AWS deck. Another put up a super resolution comparison from Google AI: bicubic, SRResNet and SRGAN set against the original, PSNR and SSIM figures printed under each frame. At five past one the seats were still full, right back to the rear wall.



