Indiana University Bloomington
Dayananda Sagar College of Engineering
-
April 2024-Present
Machine learning Engineer
Zixels LLC
Florida, USA
1. Developed an efficient algorithm using deep-CNN and Long Short-Term Memory (LSTM) and converted over 100 hours of video data for frame-by-frame analysis of 50,000+ frames to detect Autism in infants and children, improving the processing time by 20%.
2.Curated the video data using MATLAB and OpenCV to perform image analysis of facial landmarks and feature extraction for gaze estimation using Computer Vision algorithms to analyze the participant behavior in video frames.
3. Spearheaded deployment of ‘AI Tribune’ for sharing AI research, using Gatsby, resulting in a 35% increase in user engagement. -
May 2023-Present
Research Assistant
Indiana University Bloomington -
Department of SociologyIndiana, USA
Public Framing of Immigrants in Newspaper Articles:
1. Employed word embeddings using Selenium web scraping to gather 2M newspaper articles and turned the data into insights (data analysis, feature engineering, visualizations, and word embeddings) to validate the existing sociological hypothesis.
2.Built collective framing of 2M newspaper articles by fine-tuning FlanT5 LLM for zero-shot text classification achieving a precision of 95% in automatically grouping articles from different immigrant groups. -
November 2023-May 2024
Research Assistant
Indiana University Bloomington -
Department of Psychology and Brain SciencesIndiana, USA
Collaborated with the Department of Cognitive Sciences in modeling human cognition with neural networks.
1. Built a Beta-Variational Autoencoder using 1M rock images to model human visual perception resulting in a 92% correlation between the encoded features and human-collected data, demonstrating the model’s effectiveness in capturing human visual attention patterns.
2. Implemented a supervised approach using a Vision Transformer to benchmark different models in building human cognition.
-
Aug 2023-Dec 2023
Research Assistant
Indiana University Bloomington
Department of SociologyIndiana, USA
Measuring and Analyzing Boundary Rhetoric over Time:
1. Trained a BERTopic model on a corpus of 2.7 million speech data, distinguishing framing speeches of Democrats and Republicans during various presidential eras.
2. Implemented KeyBERT to extract keywords based on a predefined threshold, constructing a tailored vocabulary for the model. Achieved a coherence score of about 89% for each presidential timeframe, signifying the model's performance in topic classification. -
March 2022- August 2022
Programmer Data Analyst
Cognizant Technological Solutions
Karnataka, India
1.Developed data quality dashboards using Tableau to monitor KPIs including Test coverage, Defect density, and Test effectiveness, and identify data discrepancies, leading to a notable 15% increase in data accuracy.
2. Engineered end-to-end testing strategies with Apache Spark to optimize ETL processes, leading to a substantial 28% reduction in data processing time. Led a team of 7 interns through coding challenges and hackathons, fostering collaborative excellence. -
September 2021 - December 2021
Data Analyst
Gyaan Vivarttana Technologies
Karnataka, India
1. Led a team in utilizing machine learning techniques to analyze customer data and identify distinct segments based on their purchasing patterns. Tailored data-driven marketing strategies to each segment, resulting in a significant 15% increase in customer engagement and a remarkable 10% boost in revenue.
2. Worked on IOT-related projects - Published Temperature and humidity readings from a DHT11 sensor via MQTT with NodeMCU: Published the sensor readings to NodeRed Dashboard and extended the use of this to rain and smoke sensors to increase the use in various applications. -
March 2021 - June 2021
Machine Learning Engineer
Verzeo Technologies Ltd
Karnataka, India
1. Streamlined data analysis and forecasting using SQL queries to extract valuable trends and insights from Supermarket monthly sales data sourced via Google Cloud BigQuery, to develop and deploy the predictive ML model in Amazon SageMaker.
2. Developed an ML model for predicting human skin tone incorporating K-means Clustering and image segmentation techniques to generate masks and calculate areas to classify skin tones, deployed the model on the Google Cloud Platform.
About Me
I am a Machine Learning Engineer with over 3 years of experience in analyzing and interpreting large datasets to uncover insights and improve business performance. I have worked with healthcare and computational social science datasets, and my expertise lies in Computer Vision, Natural Language Processing, Statistical modeling, Data visualization, and Machine learning.
Proven expertise in working with multidisciplinary teams where I have analyzed data from multiple sources and turned them into actionable insights. I'm a ML Engineer at Zixels, where I'm working on gaze estimation using various CV algorithms for early detection of Autism in infants using frame by frame analysis of Video data. I work as a research assistant in computational social sciences,
focusing on racial immigration and cohesion between immigrant groups in the United States. My recent work with the Department of Psychology and Brain Sciences involves modeling human cognitive behavioral perspectives with Variational Autoencoders.
Fun fact:
I recently co-authored a paper (submitted to ACL) based on benchmarking stance detection in a zero-shot setting by fine-tuning an LLM, FlanT5-XXL model through various decoding strategies, surpassing SoTA benchmarks like RoBERTa and ChatGPT models in Semeval 2016 and Pstance datasets.
Link to the paper https://doi.org/10.48550/arXiv.2403.00236
I'm an AI tech enthusiast and love to read research papers related to multidisciplinary fields of AI.
