Joshua Choi

About Joshua Choi

I believe that data scientists are the modern-day alchemists. Data by itself is like a stone—not helpful; however, data scientists know how to turn data into gold. This imagery of turning something useless into something so valuable that it can help people, companies and countries solve meaningful problems has always seemed mesmerising to me. I was a MCDS (Master of Computational Data Science) student at Language Technologies Institute in School of Computer Science at Carnegie Mellon University. Previously, I did my undergrad at CMU, where I majored in Statistics & Machine Learning and minored in Business Analytics & Optimisation. Moreover, I was not only a 20••••22 SBVPO(Student Body Vice President of organizations), but also one of CMU\'s Quantitative Social Science Scholars of 2023 who was recognized as an individual researcher who gave the best presentation at 2023 Meeting of the Minds.I am deeply interested in pursuing projects related to applied data science, machine learning, and consulting. My working experiences include data analysis at Yuuzoo Corp(E-commerce company), SWE and ML engineering at Singapore Armed Forces, software development and data science at Behaivior, data engineering and data science at FedEx Ground. Currently, I am working as a senior DS at BNY AI HUB based in Pittsburgh. During my free time, I work as a freelance programmer/analyst/translator.https://jc-78.github.io/JC_Intro_Site/#Portfolio noteworthy Kaggle competition submission

Experience

1. Senior Data Scientist

   BNY

   Mar 2025 — Present · US

   Market Intelligence Agent •Eliza Sales Assistant (Multi-agent Lead Recommendation Engine)•Lead developer of Sanction Quality Assurance(Multi-agent reasoning system with prompt compression)*Advancing sanctions compliance by redesigning a manual QA process into agentic workflow that can emulate analyst review of screening dispositions, investigation quality and supporting evidence *LLM as Judge Research lead•Payment automation Lockbox : Multi-modal LLM prompt engineering (visual bounding box, n-image-shot prompting, modality fusion)*Improve how Eliza processes over 16 million checks for CARLAR (Courtesy Amount Recognition / Legal Amount Recognition) across 323 clients — enabling us to recover data from 83% of checks previously missed by vendor OCR system*Token-Efficient Multimodal Reasoning and LLM confidence score research lead •Eliza Bootcamp Technical Lead/ Eliza Ambassador/ Hackathon Technical Advisor: democratising the usage of AI with an impact over 1600+ employees and counting. •Eliza Spreadsheet Translation + Real-time Conversation Translation Agent* Helping APAC engineering respond faster to roughly 100 regulatory engagements by supporting translation across Mandarin, Korean and Japanese- reducing reliance on external vendors and easing time pressure on global SMEs facing tight regulatory deadlines •Lead of OpenAI AgentKit Integration with Eliza Prototyping•Mortgage agent*automating residential mortgage operations by reducing manual effort and enabling scalable processing during peak volumes.•Author of two paper publications within the first year of joining the firm*\"CompactPrompt: A Unified Pipeline for Prompt and Data Compression in LLM Workflows\" at oral presentation at the AI for Finance(AI4F) Symposium, part of ACM ICAIF 2025, as the primary author*\"A Methodology for Assessing the Risk of Metric Failure in LLMs Within the Financial Domain\" at NeurIPS 2025 Workshop on Generative AI in Finance

Education

• Carnegie Mellon University School of Computer Science

  Master's degree, MCDS

• Carnegie Mellon University

  Bachelor's degree, Statistics&Machine Learning

  2019 — 2023