Experience

Know Labs, Inc.

• United States
• Medical Equipment Manufacturing
• 1 - 100 Employee

• Head of Machine Learning and Data Science

  • Nov 2023 - Present

  Transforming non invasive medical diagnostics Transforming non invasive medical diagnostics



University of Illinois Chicago

• United States
• Higher Education
• 700 & Above Employee

• Senior Research Specialist

  • Jul 2020 - Present

  Developed a successful network that generates adversaries for multivariate time series classifiers via knowledge distillation Proposed integrating octave convolutions into time series classification models that statistically improves all DNN classifiers Developed a successful network that generates adversaries for multivariate time series classifiers via knowledge distillation Proposed integrating octave convolutions into time series classification models that statistically improves all DNN classifiers



UnitedHealthcare

• United States
• Hospitals and Health Care
• 700 & Above Employee

• Principal Machine Learning Scientist

  • Jun 2022 - Nov 2023

  • Secured 4 patents related to various aspects of machine learning and data analysis, demonstrating expertise and creativity in the field. • Devised and implemented a personalized ensemble spell check algorithm using a large language model and several Levenshtein distance based models, which achieved a 10% increase in accuracy. • Engineered an autocomplete algorithm using Trie search, resulting in a more efficient and effective search experience for users. • Operationalized and… Show more • Secured 4 patents related to various aspects of machine learning and data analysis, demonstrating expertise and creativity in the field. • Devised and implemented a personalized ensemble spell check algorithm using a large language model and several Levenshtein distance based models, which achieved a 10% increase in accuracy. • Engineered an autocomplete algorithm using Trie search, resulting in a more efficient and effective search experience for users. • Operationalized and optimized the unified search algorithm using various embeddings of large language models such as GPT, leading to a 10% improvement in recall and better search results. • Built and fine-tuned several condition prediction models using random forest, XGBoost, and DNNs, leading to a 2x increase in click-through rate and improved user engagement. • Proposed and developed an image classifier using MobileNet that identifies pills with 90% accuracy, surpassing the current state-of-the-art. • Created an innovative app that detects negative interactions between medicines, improving patient safety and healthcare outcomes. Show less • Secured 4 patents related to various aspects of machine learning and data analysis, demonstrating expertise and creativity in the field. • Devised and implemented a personalized ensemble spell check algorithm using a large language model and several Levenshtein distance based models, which achieved a 10% increase in accuracy. • Engineered an autocomplete algorithm using Trie search, resulting in a more efficient and effective search experience for users. • Operationalized and… Show more • Secured 4 patents related to various aspects of machine learning and data analysis, demonstrating expertise and creativity in the field. • Devised and implemented a personalized ensemble spell check algorithm using a large language model and several Levenshtein distance based models, which achieved a 10% increase in accuracy. • Engineered an autocomplete algorithm using Trie search, resulting in a more efficient and effective search experience for users. • Operationalized and optimized the unified search algorithm using various embeddings of large language models such as GPT, leading to a 10% improvement in recall and better search results. • Built and fine-tuned several condition prediction models using random forest, XGBoost, and DNNs, leading to a 2x increase in click-through rate and improved user engagement. • Proposed and developed an image classifier using MobileNet that identifies pills with 90% accuracy, surpassing the current state-of-the-art. • Created an innovative app that detects negative interactions between medicines, improving patient safety and healthcare outcomes. Show less



Anthem, Inc.

• United States
• Hospitals and Health Care
• 700 & Above Employee

• Senior Artificial Intelligence Machine Learning Scientist

  • Sep 2021 - Jun 2022

• Artificial Intelligence Machine Learning Scientist

  • Mar 2020 - Sep 2021

  Construct a melanoma image classifier with an AUC of 93% using an ensemble of SimCLR and EfficientNet models Create and integrate a COVID-19 nowcasting model to improve policy making using various time series data sources Implement interpretable machine learning models to identify risk factors of various users Advance time series AI solutions that are integrated in new Anthem products that potentially will revolutionize healthcare



X, the moonshot factory

• United States
• Research
• 400 - 500 Employee

• AI Resident

  • Nov 2019 - Feb 2020

  X is a moonshot factory. Our goal is to invent and launch breakthrough technologies that have the potential to solve the world's biggest problems. As an AI resident, I am developing deep learning models to help launch these breakthrough technologies. X is a moonshot factory. Our goal is to invent and launch breakthrough technologies that have the potential to solve the world's biggest problems. As an AI resident, I am developing deep learning models to help launch these breakthrough technologies.



University of Illinois Chicago

• United States
• Higher Education
• 700 & Above Employee

• Artificial Intelligence Researcher

  • Jun 2016 - Nov 2019

  Built a model to classify univariate time series using various deep learning neural network blocks, LSTM and CNN, that outperform all existing state of the art algorithms Enhanced the current univariate state of the art algorithm (LSTM-FCN) for multivariate time series classification problems by appending a squeeze and excite layer, leading it to be the current state of the art algorithm Created a White Box and Black Box Adversarial Transformation Network that generates adversaries for… Show more Built a model to classify univariate time series using various deep learning neural network blocks, LSTM and CNN, that outperform all existing state of the art algorithms Enhanced the current univariate state of the art algorithm (LSTM-FCN) for multivariate time series classification problems by appending a squeeze and excite layer, leading it to be the current state of the art algorithm Created a White Box and Black Box Adversarial Transformation Network that generates adversaries for time series classification models using TensorFlow Designed pathological voice disorder classification models using Mel-Cepstrum Vectors, LSTM-FCN, and SVM, which performed comparably with the current state of the art algorithms Proposed an Embedded Fully Convolutional Network (EFCN) to predict neurological outcomes of cardiac arrest patients in Chicago, that is being utilized to improve each cardiac arrest patient’s survivability rate

• Graduate Teaching Assistant

  • Jan 2014 - Nov 2019

  Taught ‘Engineering Statistic’ to over 100 undergraduate students where students learnt about different statistical techniques Held office hours for Engineering Statistics and Safety Engineering to help students with general questions and homework problem Assisted Engineering Quality Control by holding office hours, and grading homework assignments.

• Senior Data Scientist

  • Aug 2013 - May 2016

  Analyzed big data for model development to predict seizures of epileptic patients by applying genetic algorithm and support vector machine Constructed an expert student retention model by employing artificial neural network Created a web app for Mechanical and Industrial Engineering students to see academic progress utilizing JavaScript and R Established a Bayesian network model to infer courses student should take the following semester Built a capacity forecasting web app using… Show more Analyzed big data for model development to predict seizures of epileptic patients by applying genetic algorithm and support vector machine Constructed an expert student retention model by employing artificial neural network Created a web app for Mechanical and Industrial Engineering students to see academic progress utilizing JavaScript and R Established a Bayesian network model to infer courses student should take the following semester Built a capacity forecasting web app using Django-Python, predicting course capacity of two subsequent semesters for the College of Engineering. Developed a framework detecting optimal partial observations of various time series to decrease testing time and to reduce data purchase cost



Nokia Bell Labs

• Telecommunications
• 700 & Above Employee

• Summer Intern

  • Jun 2018 - Aug 2019

  Integrated a multi-speaker text-to-speech algorithm by using Google’s Global Style Token Tacotron Model on TensorFlow Applied a state of the art unsupervised speaker verification algorithm, developed by Google, on TensorFlow and PyTorch Utilized a speaker diarization model via speaker verification embedding and an Unbounded Interleaved-State RNN (UIS-RNN) model Presented the developed algorithms and tools to the President of Bell Labs amongst the top 5% of interns Integrated a multi-speaker text-to-speech algorithm by using Google’s Global Style Token Tacotron Model on TensorFlow Applied a state of the art unsupervised speaker verification algorithm, developed by Google, on TensorFlow and PyTorch Utilized a speaker diarization model via speaker verification embedding and an Unbounded Interleaved-State RNN (UIS-RNN) model Presented the developed algorithms and tools to the President of Bell Labs amongst the top 5% of interns