I am Ananth, a Machine Learning and Deep Learning Enthusiast. Currently, I am working as a AI Software Engineer at Delta Airlines.
Apr 2022 - Present
- Developed a Q/A model on top of OpenAI GPT 3.5 LLM using RAGs Pipeline for an internal application to improve the experience of reservation agents.
- Designed metrics for evaluating RAGs Pipeline with inspiration from OpenAI “evals” framework and TruLens. This improved the answer generation of the LLM architecture by 23%.
- Implemented project features independently - which involved design, infrastructure deployment, logic development and maintenance.
- Developed a classification algorithm for an internal application to improve travel analytics and prediction to alleviate employee travel experience.
Jun 2021 - Mar 2022
- Developed and validated complex machine learning and deep learning models on medical images (Chest X-Rays, Mammograms and Digital Hand Images) to identify racial bias.
- Developed various modules in Niffler - A DICOM Framework for Machine Learning and Processing Pipelines.
- Worked with a client to develop a Deep Learning Pipeline to detect cancers in Mammogram Images.
Oct 2020 - May 2021
- Kick-Started a Machine Learning project to cluster the customers based on their interests and market to their needs accordingly.
- Worked with a client on Data Analysis, Visualizations and automated operations to analyze the data quality. We built a unified data repository of customers to apply Machine Learning models and extract insights from the data.
- Developed applications to store, access, retrieve, analyze and build models on the data from the unified data repository.
Aug 2019 - May 2021
- Implemented a time-series forecasting model to forecast the blood-glucose levels of diabetic patients to treat hyperglycemia and hypoglycemia. The research paper has been accepted at ECAI - 2020.
- Implemented a Few-Shot Learning algorithm to fine-tune a few image triplets created from CheXpert dataset and improved pathology classification results by decreasing false-positives and false-negitives. We have experimented with usual Few-Shot Learning and Incremental Few-Shot Learning models. MarginRakingLoss is used as the loss function to implement Few-Shot Learning model. This research was performed at PLHI Lab - IUPUI under the supervision of Prof. Saptarshi Purkayastha.
- Worked on developing language models to identify and reduce the time taken to diagnose rare diseases from patient notes. This involved extracting symptoms, disease/disorders of a given patient over a period of time to narrow down the probability of the sickness being a certain disease. The research is being conducted at Data Lab - IUPUI under the esteemed guidance of Prof. Sunandan Chakraborty. We used Entity and Information Extraction techniques to extract the relevant entities from patient notes and model them accordingly.
- Worked on a POC to develop and implement a Natural Language Processing algorithm to detect and explain the presence of fake and false new statements collected from various media outlets in the US.
Dec 2017 - Jul 2019
- Developed a Machine Learning solution (classification model) on highly sensitive healthcare data to reduce the time taken by the operations team by 60% to determine the process of claim-denial in the health insurance industry.
- Implemented a Language Processing Model to process and rank resumes based on experience, education against a given job description as a part of a pilot project. The pilot project was not continued because of the discrepancies and efficiency issues in the Language Model.
2022
- Mentored a team of two students in enhancing and developing workflow modules in Niffler with Emory BMI to improve the speed and efficiency of the package usage.
2024
- Proposed projects to develop new features for Niffler. Mentored a student to develop AWANTA Framework.
2018
- Implemented a scripting mechanisum to automate the daily operations of Machine Learning Engineers and Data Scientists. The developed scripts could be controlled through text and audio.
- Won first position among 20 participating teams.
2019
- Designed and implemented a Mobile application to visualize and analyze the opioid crisis problem in Indiana. The application provides detailed analysis for each county based on the historical data of opioid usage, poverty rate, and other socio-economic factors.
- Won first place in the Mobile application Section out of 9 participating teams.
2020
- Proposed and developed tools and capabilities to detect and analyze emergency events from live streaming multimodal public safety data.
2021
- Developed python scripts to convert data from SPARC to NWB (Neurodata without Borders) format. Metadata along with demographic information and neural electrodes information is converted and stored in a CSV format for further analysis.
- +1 317-701-8138 ☎️
- Gmail