Ori Reshef

Ori Reshef did not come up through pipes and pumps, he came up through data. For two decades he built software that turns messy enterprise data into something a business can actually use, heading the product team for the customer data platform at SAP (the software that runs the back office of most of the Fortune 500) and running products at Varada, a big-data indexing startup that got acquired. As he puts it, his first love is data and his second love is nature and the planet, and he spent years looking for a way to combine the two. He found it in the unlikeliest place, the sewer.

Ori Reshef met Kando's founder, Ari Goldfarb, eight or nine years before he joined, and the pitch stuck with him: the wastewater flowing under a city is one of the richest, hardest-to-mine datasets on the planet, and almost nobody was reading it. Kando puts sensors in the sewer network, the middle ground between the city upstream and the treatment plant downstream, to find where pollution is coming from in real time and get the source to change its behaviour before it ever reaches the plant. Ori Reshef consulted as a friend for a couple of years because he believed in the mission, and at one point decided to stop consulting and commit.

Ori Reshef's headline contribution at Kando is STREAMi, a conversational AI assistant for the water utility, and it speaks to exactly the kind of digital water shift the sector keeps promising itself. The problem it solves is a real one: utility executives told him they understood that wastewater quality matters, but they were buried under too much data and a shrinking workforce, and they simply did not have the time to dig through it. STREAMi lets them ask, in plain language, what is happening in their network and what to do about it. The clever, and honest, part is how he built it. A 50-person company cannot train its own large language model (an LLM, the kind of AI behind ChatGPT) the way Google or Microsoft can, so STREAMi sits on top of existing models and feeds them Kando's own sensor data and expertise, which is the only practical way a company that size ships generative AI in 2024.

Ori Reshef is refreshingly unromantic about the AI itself. His line is that the chat is the easy part and the data underneath is the whole game: the sensors, the machine-learning models, the anomaly detection. A convincing assistant built on bad data, he warns, just gives you confident nonsense, garbage in and garbage out, and the real risk is that you follow it. He is equally measured about where AI stops, arguing the human operator should be augmented and not replaced, because creativity and empathy are still things the machines cannot do. It is the view of someone who has watched a lot of data hype come and go.

“The most unexpected partnership was between the two poles of product house and customer experts. This partnership, this friendship between the two parts really improved our product roadmap in ways I cannot imagine. The product is everybody's, especially those who sit with the customers.”

Ori Reshef has since moved on from Kando, having co-founded a new venture and, as of early 2026, taken a data leadership role at the trading platform eToro, but the STREAMi episode is a clean snapshot of how he thinks: point world-class data craft at the messiest, most-ignored dataset you can find, and let the people who own the problem finally talk to it.