Help & Supportcareers@gravityer.com
Principal Applied Scientist at Amazon
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Hiring to solve interesting problems related in customer trust and partner support (CTPS) domain at Amazon. Working to push boundaries of automation in risk management systems and tackle high-scale adversarial cyber attacks from seasoned bad actors.
University of Minnesota
Masters, Computer Science
January 1, 2012 – January 1, 2014
Amazon
Principal Applied Scientist
April 1, 2022 – Present
Amazon
Senior Machine Learning Scientist
April 1, 2018 – April 1, 2022
Amazon
Machine Learning Scientist II
April 1, 2016 – March 1, 2018
Amazon
Machine Learning Scientist
February 1, 2014 – March 1, 2016
Amazon
Research Scientist Intern
June 1, 2013 – August 1, 2013
Greater Seattle Area
University of Minnesota
Graduate Research Assistant
January 1, 2013 – April 1, 2013
NVIDIA
System Software Engineer
June 1, 2011 – July 1, 2012
Multi-label classication on a dataset having 6 million records
October 1, 2013 – December 1, 2013
We built machine learning models to assign correct tags to StackOverflow questions. We used two approaches- random forests and classifier chains. The key challenge was data munging / cleaning the dirty data.
Music Anywhere – Kinect Based Virtual Musical Instrument
October 1, 2012 – Present
Developed an application capable of sensing any suitable object in the field of view of the Kinect sensor and use it as a musical instrument. Performed image processing on surrounding objects to detect their shapes and to classify them into two categories. Upon touching those objects, the application produced sound of appropriate frequency. Explored open source tools like OpenNI, Kinect driver, SkelTrac library, OpenAL, PortAudio and OpenCV for the project.
Volatile Split Cache Architecture: An architectural solution to incorporate STT-RAM in Lower Level of Caches.
August 1, 2012 – December 1, 2012
As the technology scales down the transistors in SRAM become more leaky leading to extremely high leakage power. One of the leading alternatives to replace SRAM is STT-RAM. STT-RAM is a type of MRAM that uses spin transfer current to toggle between states. MTJs have a very low leakage current and comparable read latencies to that of SRAM. STT-RAMs however suffer from very high write latencies and high write dynamic energies. Although STT-RAM is categorized as a non-volatile memory, we exploit the unexplored volatility property of STT-RAMs’ to obtain a higher performance and better power numbers when compared to a complete non-volatile STT-RAM based cache. In this paper, we firstly explore the possibility of using STT-RAM as a direct replacement to SRAM on various levels of cache hierarchy followed by proposing a new hybrid architecture (Volatile Split Cache Architecture - VSCA) which exploits the volatility property to achieve enhanced power numbers along with better performance when compared to a fully non volatile STT-RAM based cache. Direct replacement of SRAM with STT-RAM in the L-2 cache saves 87% power over the SRAM (for a write pulse width of 10 seconds, the power improvement would be better for a larger pulse width but at the cost of performance), whereas when we exploit the volatility property of the STT-RAM using our VSCA architecture, we further save about 12.5% of the power as compared to the pure STT-RAM replacement; and at the same time achieving better performance numbers (30% improvement in read latency 50% improvement in write latency). For exploiting the volatility property it is critical to find out the optimum value of retention time of the MTJ. For doing so we firstly obtain the inter-write latencies per cache block using the GEM5 simulator for some of the SPLASH benchmarks and decide upon the retention time based on this application characteristics. Our analysis shows we can readily replace SRAM by STT-RAM saving power.
Robocon
December 1, 2007 – March 1, 2011
Asia's biggest Robotics Competition
Cultural Fit Analysis
The candidate has a strong background in research and applied science, which aligns well with roles requiring innovation and problem-solving. The project diversity, ranging from hardware architecture to robotics and machine learning, indicates a broad intellectual curiosity. However, the target role is 'Data Analyst', which is a significant shift from 'Principal Applied Scientist' focusing on LLMs. While the candidate has experience with data (e.g., multi-label classification), the depth of experience specifically in data analysis tools, reporting, and business intelligence is not explicitly detailed, which could impact cultural fit for a pure Data Analyst role.
Soft Skills & Operational Fit
The candidate's career progression at Amazon suggests strong operational fit and the ability to handle complex projects. The project descriptions, while technical, lack explicit details on collaboration or leadership, making it difficult to assess soft skills beyond what is implied by a Principal role.