Experience

• 

  Qatar computing research institute

  May 2018 - Jul 2018

  Data analytics intern

  • Worked on a research project to develop a novel vector representation model for an avionics system consisting of two types of fault messages: maintenance message and flight deck effects.• This was implemented using Google's word2vec and the performance was compared against the existing results from the logistic regression model.

• 

  Innov4sight health and biomedical systems private limited

  Feb 2019 - Jul 2019

  Data science intern

  • Worked on identifying adverse event reactions from product feedback verbatim by creating a multi label text classification model to classify the types of products used by customers based on the product feedback/adverse event reactions. • Obtained an efficient performance of 89% by leveraging word embeddings and RNN-LSTM models.

• 

  Amrita vishwa vidyapeetham

  May 2019 - Mar 2020

  Undergraduate research assistant
• 

  Omdena

  Sept 2019 - Mar 2020

  • Worked with a diverse team of 20 AI collaborators from across the world to provide a solution for identifying sexual abuse at workplaces and online by collaborating with Zero Abuse Project. • Led the data wrangling team to perform web scraping of ~1M cybercrime chat-log data and built dashboards using Plotly for insights• Effectively communicated data-driven insights to improve decision-making and collaborated with cross-functional teams• Spearheaded the detection of online abuse crimes using unsupervised ML and language models to achieve a performance of 86%• Interacted with stakeholders and project managers to deliver data visualizations and data-driven insights Show less Omdena is a global platform that unites mission-driven organizations with AI engineers, data scientists, and domain experts from diverse backgrounds, all working together to harness the power of AI for meaningful causes.One of our recent partnerships was with World Resources, where we embarked on a project aimed at resolving environmental land conflicts in India and linking them with land restoration policies. The collaborative effort involved 30 AI Enthusiasts hailing from 30 countries, creating a truly international and multidisciplinary team.• In partnership with World Resources, the project involved in resolving environmental land conflicts in India and connecting them with policies for land restoration. This was done in the collaboration with 30 AI Enthusiasts from across 30 countries. • Involved in scraping articles from news events, matching government policies to conflict new to get a better understanding of policy gaps.• performed coreference resolution on news text data using Spacy and Neural-coref and was actively involved in data annotation of news articles which was necessary to classify conflict news articles. Show less

  • Lead ML Engineer

    Nov 2019 - Mar 2020

  • Junior ML Engineer

    Sept 2019 - Nov 2019
• 

  Sierra odc private limited, india

  Apr 2020 - Jul 2020

  Data science intern

  • Performed data exploration and quantitative analysis on time-series data to obtain and communicate various data-driven inferences• Optimized energy consumption forecast of a building utilizing time-series statistical tests and enhanced the model performance by 15%.• Leveraged time series models (ARIMA, Prophet), and LSTMs to predict future electricity consumption and solar energy production for the building• Developed a user-friendly Flask Application with predictive analysis which was hosted on Heroku. Show less

• 

  Innov4sight health and biomedical systems private limited

  Aug 2020 - Aug 2021

  Machine learning researcher

  - Developed a clinical narrative information extraction tool to identify adverse event reactions to products/medications used by customers/patients by utilizing Python and NLP libraries- Obtained a performance of 85 % by implementing Named Entity Recognition for extracting data from published articles- Performed analysis in Python and developed a multi-label classification pipeline of the types of products used based on product feedback.- Achieved an efficient performance of 89% by utilizing language models, and transformer BERT models. Show less

• 

  Khoury college of computer sciences

  Sept 2021 - May 2023

  - Mentored undergraduate students to enhance their abilities in Python programming, data analysis, hypothesis testing, and data science core concepts.- Coordinated with the professor and fellow teaching assistants to lead recitations, prepare coursework and structure of teaching.- Held weekly office hours to assist students with data science lab assignments and homework. Graduate Teaching Assistant for course DS3000-Foundations of Data Science under Prof. Sophine Clachar

  • Graduate Teaching Assistant

    Dec 2021 - May 2023

  • Graduate Teaching Assistant

    Sept 2021 - Dec 2021
• 

  Reboot rx

  Jul 2022 - Dec 2022

  Data scientist co-op

  - Created a workflow pipeline to summarize the findings from the manually annotated training data required to create high-quality datasets required for improving the performance of the end-to-end pipeline models used for detecting non-generic cancer drugs.- Generated labeling functions using high-quality manually annotated data by incorporating a rule-based approach to propagate labels on unlabeled training data and test the coverage and correctness of labels.- Improved data quality by 15% through comprehensive analysis of 1,500+ annotated training samples and implementation of advanced computational methods like Influence Functions on AWS.- Boosted workflow efficiency by 20% by developing 10 interactive Dash dashboards for data visualization, enhancing team decision-making processes.- Collaborated with clinical scientists to analyze over 1,000 scientific publications and clinical trial datasets, resulting in a 10% improvement in BERT model performance for medical text classification.- Enhanced relevance classification accuracy to 85% by optimizing BERT models with targeted text inputs and augmented training data from unique article spans.- Conducted 5 research experiments using various BERT models (PubMedBERT, SciBERT, BioBERT) to identify non-generic cancer drugs, improving end-to-end pipeline performance Show less

• 

  Khoury college of computer sciences

  Sept 2023 - Dec 2023

  Graduate teaching assistant

  Graduate Teaching Assistant for CS4100 Artificial Intelligence

• 

  Narwal

  Dec 2023 - now

  Job Prioritization Scoring and Retraining ProjectWorked on a project with a healthcare staffing client to address inefficiencies in prioritizing a high volume of job requests, which impacted their ability to capitalize on critical opportunities. Recruiters were overwhelmed by the manual process, struggling to identify high-priority tasks, which led to missed revenue opportunities. - Built and deployed a predictive model using LightGBM and advanced statistical methods, improving job request prioritization by 30%, which contributed to a ~2M revenue increase during the project lifecycle.- Deployed the solution in production, cutting recruiter workload by 50% (from 10 to 5 hours per week) and reducing job prioritization time by 40%.- Established performance monitoring and retraining workflows to mitigate data drift, achieving a 15% improvement in prediction accuracy.Advanced Multilingual PDF Analyzer Bot Using LLM- Collaborated with cross-functional product teams and leveraged Mistral AI to develop a multilingual PDF analysis bot, increasing financial data analysis productivity by 40% and effectively communicating insights to stakeholders.- Designed and implemented a conversational AI interface for the PDF analyzer using RetrievalAugmented Generation (RAG), enabling natural language queries and improving user engagement by 35%.- Automated data extraction from structured and unstructured sources, saving 25+ hours of manual effort and enhancing data accessibility with advanced PDF analysis techniques. Show less In my role as a Data Science Intern at Narwal, I worked with the cybersecurity team of a large client to automate the detection of abnormal Kerberos attack patterns, addressing their entirely manual and time-consuming process that required approximately 5 hours per case using MLFlow in Databricks, reducing manual analysis time by 5 hours per case. - Engineered ETL pipelines and applied custom data models to process and analyze 4.3 million Active Directory event logs.- Achieved 96% AUC-ROC score by experimenting with supervised and unsupervised models.- Designed and deployed an end-to-end MLOps pipeline with continuous KPI monitoring and alerting systems. Show less

  • Data Scientist

    Mar 2024 - now

  • Data Science Intern

    Dec 2023 - now
• 

  Fis

  Aug 2024 - now

  Data scientist

  - Collaborated on financial data migration from BDA platform to CDP, enhancing real-time analytics capabilities.- Automated HTML report generation using R Markdown, reducing report preparation time by 40%.- Conducted in-depth analysis of raw financial data, boosting report accuracy by 25% through rigorous cross-referencing.