"Shahbazi has competed in 20 Kaggle events, coming second five times and finishing in the top 15 another five times. He finished in the top five per cent of challengers on contests for everything from a Grupo Bimbo (the Mexican bakery multinational that owns Dempsters, Hostess, Vachon brands in Canada) inventory algorithm, a Home Depot online search relevancy challenge, a drug store sales volume prediction engine, a music recommendation system and even the 2017 Data Bowl that looked at lung-cancer detection".

"Zillow has slowly improved its Zestimate from a median error rate of 14 per cent when it started in 2006 to 5.7 per cent when the contest began in mid-2017. It's now down to 4.5 per cent nationally (it's higher in some cities and lower in others), and once the winners' tweaks to the algorithm are incorporated, the company expects the error rate to dip to about 4 per cent".

"Jordan Meyer of the United States, Chahhou Mohamed of Morocco, and Nima Shahbazi of Canada bested more than 3,800 teams representing 91 countries with an algorithm that beat Zillow's benchmark model (evaluated against real-time home sales between August and October 2018) by approximately 13 percent."

"It's amazing to know that millions of people will benefit from our ideas," said Shahbazi, who competed as a team with Mohamed against Meyer in the contest's initial qualifier, but who decided to join forces with Meyer for the final round. "We brought every novel idea we could to our code and kept experimenting. For every idea that worked, there were a hundred that didn't work. But we kept going."

On average, Zillow said, the Zestimate is $10,000 off the actual sale price for a median-priced home of about $223,900, and the information gleaned from the Zillow prize winnings could shave $1,300 off that discrepancy. It also moves the Zestimate's national median error rate below 4%.

"Drug store giant challenged Kagglers to forecast 6 weeks of daily sales for 1,115 stores located across Germany. The competition attracted 3,738 data scientists, making it our second most popular competition by participants ever.
Nima Shahbazi took second place in the competition, using his background in data mining to gain an edge. By fully exploring and understanding the dataset, Nima was able to engineer features that many participants overlooked."

"Asked to search for signal in financial markets data with limited hardware and computational time, this competition attracted over 2000 competitors. In this winners' interview, 2nd place winners' Nima and Chahhou describe how paying close attention to unreliable engineered features was important to building a successful model."

"The winning team leveraged the power of NVIDIA TITAN Xp GPUs for both training and inference. For software, the team used the cuDNN-accelerated Keras, and TensorFlow deep learning frameworks".