Kiran Ramnath

I am a Data Scientist at Salesforce, where I work on problems in natural language processing and machine learning.

I received my MS in Electrical and Computer Engineering from University of Illinois, Urbana Champaign in May, 2021 where I worked on Fact-based Visual Question Answering using Knowledge Graph Embeddings advised by Prof. Mark Hasegawa-Johnson. Prior to that, I worked as a Data Science Consultant with PricewaterhouseCoopers' US Advisory in Mumbai from 2016-19, advising clients on realizing business goals through data-driven insights. I received my Bachelor's in Electrical and Electronics Engineering from BITS, Pilani in 2016.

Email  /  CV  /  Bio  /  Google Scholar  /  Github  /  Linkedin

Fun fact: My academic family tree: me - Hasegawa-Johnson-Stevens - Beranek - Hunt - Chaffee - Pierce - Macfarlane - Tait - Hopkins - Sedgwick - Jones - Postlethwaite - Whisson - Taylor - Smith - Cotes - Newton.

I'm interested in machine learning, natural language processing, computer vision, and knowledge graphs - particularly in multi-modal applications of deep learning.

We use KG Embeddings to help the network reason over incomplete knowledge graphs to answer fact-based visual questions.

MS Thesis

We perform question-answering directly using speech signals as input over images and knowledge graphs.

In my past life, I performed undergraduate research in VLSI. We demonstrate the benefits of forward-biasing the body of subthreshold op-amps which improves gain and reduces noise in the device.

We trained a deep-learning based policy for a robotic wireless access point to provide better indoor WiFi signal to a receiving device. The transmitter was strapped onto the end of a 5 DoF robotic arm (Locobot) and the receiver was placed in various locations in the house. The policy predicts which a series of actions that would result in overall betterment of the receiving signal for the bot.

Working with Hyperforce Development Platform Support team that provides customer support to developers on CI/CD as Salesforce migrates to Public Cloud. Currently working on NLP use-cases such as search improvement to help reduce support burden by returning relevant technical documentation on Confluence to customers. Work also includes metrics generation and data visualization to monitor bot performance and suggest process improvements.

Worked with Capacity Planning team to predict Salesforce’s global server requirements in 3-5 years. Performed demand segmentation and built 30+ forecasting models using FB Prophet. Automated periodic analysis by dockerizing data pipelines.

Worked across sectors like Finance, Healthcare, and Technology implementing data science use-cases. Project areas varied across customer analytics, market research, and company-wide analytics transformation. Modelling techniques spanned linear / logistic regression, decision trees, topic-modelling, random forest, bayesian networks, and deep learning.

Digital Product School is an intense 3-month fellowship program conducted by Center for business and Innovation at TU Munich, aimed at grooming the next generation of product makers to thrive in cross functional teams in a lean startup environment, with a prime focus on user-centered design. I worked as an AI Engineer partnering with BMW and Nokia to build a solution to help cities deal better with traffic flow. My work involved conducting user interviews, ideating and managing product feature pipelines, building back-end infrastructure, and conducting feasibility studies using Agent-based simulation.

Worked on developing and managing content, design and operations of a workplace optimization web portal that allowed managers to remotely track and monitor a new hire's progress. Conceptualized the entire pipeline and implemented frontend and backend using Python's Django framework, along with unit testing using Selenium and continuous deployment using Jenkins Improved ramping up time and efficiency for 100 new hires during the course of the internship

Worked as a consultant data scientist contracted by Dalberg Advisors, India working along with Global Fund to End Modern Slavery (GFEMS). Helped devise their online sex-trafficking deterrence strategy by scraping ~100,000 ads from several websites and analysing top locations, suppliers, etc.

SP-Jain Institute of Management and Research, June 2019

Future of Privacy Forum, Jan 2021

Many thanks to Jon Barron's source code on which this website is based - site, source code