Ankush Chaudhari Email & Phone Number

New York, Ny, Us

New York, New York, United States

New York, New York, United States

Working in Dr. Narges Razavian’s team to design and pre-train 124M Transformer model on Electronic Health Records (EHR) for early disease detectionConducting pre-training experiments using 4 billion tokens of EHR data from 2M+ patients over 10 years

New York, United States

New York, New York, United States

San Francisco Bay Area

Working on Research and Development of Generative AI-based tool to enable context-based analysis of multimodal data.Designed scalable unstructured data processing systems, combining fine-tuned open-source LLMs and OpenAI GPT models to generate domain-specific insights. Achieved a 72% average user satisfaction rate.Engineered custom chunking, content retrieval, map-reduce, and self-consistency mechanisms to tackle GPT-3.5-16K limitations, enhancing the platform and reducing costs by up to 70% by avoiding 10X higher cost of GPT-4 adoption.

San Francisco Bay Area

Built optimized ETL pipelines using Airflow on AWS to integrate 5 different data sources handling 50M+ records/day. Designed and built a Leads Scoring internal tool that uses raw data points of over 1.5M restaurants to compute the score of their probability of conversion to Snackpass partners based on features such as orders pattern, online reviews across 18 sources, menu, etc., saving the sales teams’ approx. 3000 worker-hours/year.

Spearheaded R&D of patent-pending technology which generates crowd analytics for large areas by using ML to analyze radio wave patterns emitted by smartphones with up to 85% accuracy deployed on cost-effective portable hardware. Developed a CNN-based client menu digitization solution which reduced onboarding time from 2 hrs. to 10 min. Developed an XGBoost regression-based order preparation time estimator model trained on features such as concurrent open orders count and items specifications, current order items specifications, historical trends of the store, etc. to estimate the time duration required to complete the current order. Achieved an RMSE of under 2 min.Performed chi-squared tests on the model's results to determine its statistical significance.

Developed a CNN-based crowd analysis application to detect queues and estimate wait times with an RMSE of 2 min. Optimized deep learning processes to support GPU-based inference, which increased the throughput by 10X.

India

Worked under the guidance of Dr. Santhi V. on the SAR Satellite Image Analysis research project, proposed and sponsored by the Indian Space Research Organization (ISRO). Developed the sidelobe suppression algorithm for lossless enhancement of 16-bit satellite images using Python. Formulated a novel approach that significantly improved the quality of results and reduced the processing time from 8 hrs. to 15 min. for gigapixel resolution SAR images compared to the existing solution.

Vellore, Tamil Nadu

Vellore, Tamil Nadu, India

Responsible for the overall functioning of the student chapter.Received Presidential Award for Outstanding Student Chapter globally to be in the top 5% of all SPE student chapters across the globe.Led a team of around 35 members to work on various technical projects and host flagship events at our university.Supervised the operations of SPE's social initiative SPE Cares in Vellore.

Vellore Area, India

Hyderabad Area, India

I was the only one selected of all interns during my tenure to work along with the data science team on one of the company's core projects for a huge client.Scrapped data; cleaned it; trained a convolutional neural network on the cloud to get >90% accuracy and integrated it with the existing web app.Developed an image-based fashion product search solution by training a CNN to extract image features and cross-referencing them with the products database which reduced user search time by 40% compared to text-based search.

Vit University

Vellore Area, India

Master'S Degree, Data Science, 3.91 / 4.00

Coursework: Natural Language Processing and Representational Learning, Machine Learning, Causal Inference, Big Data, Probability.

Bachelor Of Technology - Btech, Computer Science, 3.97 / 4.00