Current position: Lead scientist at GE Global Research
As a child, Subrahmaniam had an unusual combination of interests. He was interested in mathematics as well as music. While he was fond of reading books, he liked forecasting the weather and playing the veena. He grew up to become of one the youngest and finest statisticians and data scientists in the country.
Twelve years ago, when he had to decide the course of his career, his mother encouraged him to take up the field which he so passionately adored. “My mother has always believed that to really become successful, you need to do what you are passionate about,” says Subrahmaniam.
So, at that time, when people usually opted to study for a medical or engineering degree, he went to the Indian Statistical Institute, Kolkata.
After completing his studies in 2008, he joined the global research team at the US-based products and services firm General Electric Co., where he had interned in 2005. His forte is statistical modelling for weather forecasting to solve industrial problems.
In the early stages of his career, he developed mathematical models and algorithms that added significant value to diverse industrial applications. He worked with two more researchers to develop a new machine-learning algorithm which analysed historical weather data in the Indian subcontinent to identify regions that have similar rainfall patterns. His research led to the discovery of a new monsoon region called the Homogenous Indian Monsoon Region, which later improved monsoon forecasting over large areas of northern and eastern India. His work was published in the Journal of the Royal Meteorological Society (2009).
In 2010, Subrahmaniam invented an automated mixture analysis algorithm for identifying unknown substances based on its Raman spectra. This method was later granted a patent. “It was used by GE’s Homeland Protection business to identify unknown mixtures such as narcotics or drugs correctly,” says Subrahmaniam.
Though identification of mixtures from spectral data is a well-studied topic, Subrahmaniam developed the first ever method based on partial correlation to generate probabilities on all the identified mixture components. His work on this was published in the Journal of Applied Spectroscopy in 2010.
Over the last seven years, Subrahmaniam has been involved in creating a software system consists of self-learning algorithms, which analyse historical weather patterns at a location and predict the most repeatable patterns. It compares that information with the requirements for its clients, such as setting up a wind farm at that location, to come up with the best model to do so. The method has been adopted by GE’s wind business for setting up wind turbines.
Subrahmaniam has been involved in industry applications related to wind farming, power grids, transport and aviation.
“Over a period of time, the system that we have created has demonstrated verifiable results,” he says.
Subrahmaniam is now experimenting with deep learning, a new field in machine learning, which uses artificial neural networks to learn levels of representation and abstraction that make sense of data such as images, sound, and text, in a way the human brain learns and understands it.
“We are using new analytical models and deep learning for analysing historical weather data to increase the accuracy of our predictions,” he says. “We are also looking to use this model to expand our horizons and solve problems across various industries.”
Source : Google, LiveMint.
To buy inspiring books, individual counselling and group seminars or to advertise on this blog or support us to spread positive inspiring news, please contact Sudhir Pai Training Systems on 9820073993 (Mumbai - India). We are looking forward to meet you.
