• In Python and SQL, developed a novel classifier using ultrasonic time series data to detect 50+ false negatives weekly, reducing errors by 15% in Segment Anything Model for real-time object detection for food spillage.• Collaborated with business stakeholders to pinpoint and mitigate the root causes of false negatives, leading to strategies that are projected to save $1.3 million and reduce product spillage across conveyors by 32.5%.• Built ETL processes on SAP data using SQL/Python… Show more • In Python and SQL, developed a novel classifier using ultrasonic time series data to detect 50+ false negatives weekly, reducing errors by 15% in Segment Anything Model for real-time object detection for food spillage.• Collaborated with business stakeholders to pinpoint and mitigate the root causes of false negatives, leading to strategies that are projected to save $1.3 million and reduce product spillage across conveyors by 32.5%.• Built ETL processes on SAP data using SQL/Python to identify inefficiencies with KNN/network graphs. Saved 2.2M in cost Show less

• Developed an LLM chatbot (Mistral 7 B) recommender that assist financial portfolio managers for Accenture’s clients. Used Python with Retrieval Augmented Generation to pull unstructured/structured financial data with Text2SQL Llama index• Improved perplexity by 20% via fine-tuning, reinforcement learning (RLHF) & prompt engineering. Deployed with AWS

•Increased customer satisfaction by 12% using scikit-learn (PCA, Elastic Net, Random Forests) and XGBoost for priceoptimization and LLMs/topic modeling for sentiment analysis with MLflow• Created a LLAMA2 recommender system that generates operational technology (OT) cybersecurity solutions.• Identified 20+ potential vulnerabilities in OT cybersecurity by detecting 1000+ anomalies using auto-encoders and clustering.• Utilized a probabilistic record linkage model on 6 million… Show more •Increased customer satisfaction by 12% using scikit-learn (PCA, Elastic Net, Random Forests) and XGBoost for priceoptimization and LLMs/topic modeling for sentiment analysis with MLflow• Created a LLAMA2 recommender system that generates operational technology (OT) cybersecurity solutions.• Identified 20+ potential vulnerabilities in OT cybersecurity by detecting 1000+ anomalies using auto-encoders and clustering.• Utilized a probabilistic record linkage model on 6 million rows, identifying 4,000+ potential fraudulent relationships between vendors and customers. Deployed an interactive visualization using Tableau and DBeaver for compliance review to build a predictive fraud model, achieving a 24-fold reduction in fraud identification times. Show less

• Web-scraped website reviews of 900k documentaries to create a benchmark that quantifies the social influence of documentaries.• Utilized text network, n-gram frequency, and topic modeling to generate a metric that quantifies 40% of the viewers as impacted.• Fine-tuned BERT to classify documentary reviews into impactful and non-impactful categories to assess viewer engagement.• Compared 3 data-mining methods to optimize runtime by 20% and reduced matrix sparsity by 30% to resolve… Show more • Web-scraped website reviews of 900k documentaries to create a benchmark that quantifies the social influence of documentaries.• Utilized text network, n-gram frequency, and topic modeling to generate a metric that quantifies 40% of the viewers as impacted.• Fine-tuned BERT to classify documentary reviews into impactful and non-impactful categories to assess viewer engagement.• Compared 3 data-mining methods to optimize runtime by 20% and reduced matrix sparsity by 30% to resolve memory constraints. Show less

• Examined the racialization of American public opinion on China by wrangling 30GB of 24 different broadcast television political advertisement datasets aired on the 210 media markets.• Converted 30,000 campaign video data to raw text using IBM’s Speech-To-Text API with Python.• Conducted descriptive statistics, visualization, bivariate regression, Structural Topic Modeling, and Sentiment Analysis utilizing R and Tableau.• Found sentiment discrepancies between China and non-China… Show more • Examined the racialization of American public opinion on China by wrangling 30GB of 24 different broadcast television political advertisement datasets aired on the 210 media markets.• Converted 30,000 campaign video data to raw text using IBM’s Speech-To-Text API with Python.• Conducted descriptive statistics, visualization, bivariate regression, Structural Topic Modeling, and Sentiment Analysis utilizing R and Tableau.• Found sentiment discrepancies between China and non-China related advertisements for both Democrats and Republicans over multiple types of elections.• Highlighted 4 key trends from text analysis and proposed 5 actionable suggestions for stakeholders. Show less

• Conducted primary market research and data analysis to deliver client outreach initiatives.• Executed situation, market opportunity, Facebook advertisement data, and demographic data analysis with Python and Tableau to prepare a go-to-market strategy for a concert rewards startup company.• Identified approximately 5,000 new serviceable obtainable market by wrangling Colorado demographic data with Python and Tableau.

• Assisted in managing 9 different Online Banking (OLB) web and mobile (IOS and Android) feature integration and software development projects using Agile methods from development to production stage. • Utilized capacity planning and portfolio management software to increase portfolio health KPIs by 15%. • Drafted and maintained 19 project discovery and charter documents to define strategic alignment and cost estimates. • Collaborated with 5 team members to prepare a customer… Show more • Assisted in managing 9 different Online Banking (OLB) web and mobile (IOS and Android) feature integration and software development projects using Agile methods from development to production stage. • Utilized capacity planning and portfolio management software to increase portfolio health KPIs by 15%. • Drafted and maintained 19 project discovery and charter documents to define strategic alignment and cost estimates. • Collaborated with 5 team members to prepare a customer acquisition strategy with the goal of increasing retail bank population by 20%. Show less