Data Science and Investment Research

Technical Skills: TensorFlow,NumPy, Pandas, Scikit-learn, R, SQL/NoSQL, Snowflake, SPSS, JMP/JSL, VBA, Hypothesis Testing, Regression Analysis, Experimental Design, PowerBI, Exploratory Data Analysis, Classification, LLMs

Education

• Philosophy, BA
• Biology, BS
• HuggingFace: https://huggingface.co/JuIm

Work Experience

Quantitative and Fundamental Fixed Income Researcher @ Conning Inc. (August 2022 - Present)

• Built ensemble AI credit rating models (MLPs) using TensorFlow to quantify fundamental credit risk, enabling faster and more precise investment decisions as well as potentially expanding coverage
• Architected a proof-of-concept ETL pipeline using Snowflake and Snowpark to ingest market data and generate actionable trading signals, resulting in broader market coverage and improved price discovery
• Created a proof-of-concept AI pipeline using Llama 2 and AbstractCRE, streamlining research and amplifying productivity by generating credit and ESG comments
• Prepared and presented detailed investment recommendations, memos, credit comments, and sector reports for the firm’s strategy meetings
• Reviewed and interpreted complex legal documentation such as bond indentures and credit agreements to understand structures, covenants, and associated risks
• Conducted investment research, financial analysis, and wrote in-depth reports on $7B in Municipals encompassing high yield, structured, and investment grade debt
• Performed analyses on financial and ESG datasets to identify statistically significant trading signals and created portfolio construction frameworks to evaluate signals

Data Analyst @ Boehringer Ingelheim (March 2022 - August 2022)

• Provided ongoing maintenance, support, and enhanced functionality for a globally used data quality assurance tool in Excel/Access VBA to ensure application reliability and a smooth user experience
• Streamlined operational workflows through task automation using Excel VBA to drive efficiency and better utilization of resources

Research Experience

Student Researcher @ UConn Health – Behavioural Sciences

• Investigated the relationships between facial flushing after the consumption of alcohol and various biometrics such as BP and HbA1c
• Research resulted in a peer-reviewed publication

Student Researcher @ Quinnipiac University – Environmental Sciences

• Investigated the effects of ‘forever chemicals’ on HEK 293 cell growth

Student Researcher @ Yale School of Medicine – Neuropsychiatry

• Researched the effects of flavorants and nicotine concentration on nicotine intake in rats via intra-oral and intra-venous self administration
• Work resulted in several peer-reviewed publications

Completed Data Science and Artifical Intelligence Projects

ProGemma - A Gemma 2 LLM Pre-Trained on Amino Acid Sequences of Length 0 - 512

• Pre-training a custom configuration (275M) of the Gemma 2 LLM to generate novel amino acid sequences on a letter-by-letter basis, a key application in protein engineering
• Built a custom BPE tokenizer to ensure efficient model training
• Preliminary evaluation shows perplexity and AlphaFold 3 scores are on par with the widely adopted ProtGPT2 model
• Model is being continually pushed to HuggingFace when training checkpoints are reached

ViT_Skin_Cancer – Fine Tuned ViT on Skin Cancer Images

• Fine-tuned a ViT (google/vit-base-patch16-224) on a collection of benign and cancerous skin cancer photographs
• Open-sourced the model on Hugging Face for public use and commentary

ViT_Breast_Cancer – Fine Tuned ViT on Breast Cancer Microscope Images

• Fine-tuned a ViT (google/vit-base-patch16-224) on a collection of approximately 7000 images of benign and cancerous breast tissue samples
• Achieved strong performance with a loss of 0.007 in classifying malignant and benign tissues, potentially aiding in more accurate diagnoses
• Open-sourced the model on Hugging Face for public use and commentary

SMILES_BERT – A RoBERTa model Fine-Tuned on Chemical SMILES

• Trained a custom configuration of a RoBERTa model on a dataset of chemical SMILES strong to bridge the gap between natural language processing and drug discovery
• Created and trained a custom Byte Pair Encoding (BPE) tokenizer to better handle the unique vocabulary, resulting in the elimination of ‘ ’tokens
• Customized the configuration of the base model to achieve a better compromise between available compute and model performance
• Open-sourced the model on HuggingFace, achieving 3000+ downloads showing a contribution to the advancement of the computational biology/chemistry research

Fixed Income Portfolio Optimizer

• Built a portfolio optimizer using the PuLP library in Python for fixed income instruments that maximizes portfolio yield for a given ratings, convexity, and duration target
• The optimizer supplements discretionary portfolio construction with data-driven portfolio insights, which allows for more guided and precise decision making
• Allows for a more streamlined implementation of specific portfolio strategies such as barbells, bullets, or ladders

Monte Carlo DCF Framework

• Built a DCF framework in Python that utilizes Monte Carlo methods to simulate valuations given normally distributed EBITDA Margins, Revenue Growth, and Net Working Capital

Projects to be Completed by Year End 2024

GemmaRNA – Controlled ncRNA Sequence Generation

• Pre-training a custom configuration of Google’s Gemma 2 LLM on ncRNA sequences and GO annotations to make controlled generation based on function, location, and process

Publications

Citroen, K. & Immanuel, J. (2024). State of the States: Municipal Bond Credit Perspectives [White paper]. Conning Inc. https://www.conning.com/-/media/marketingsite/documents/state-of-the-states/state-of-the-states-2024.pdf

Citroen, K. & Immanuel, J. (2023). State of the States: Municipal Bond Credit Perspectives [White paper]. Conning Inc. https://www.conning.com/-/media/marketingsite/documents/state-of-the-states/state-of-the-states—2023.pdf

Feinn, R., Bermudez-Millan, A., Berthold, S. M., Buckley, T., Immanuel, J., Fraser-King, L., Horn, I. S., Keuky, L., Kong, S., Kuoch, T., O’Donnell, S., Read, J. P., Scully, M., & Wagner, J. (2021). Relationship of alcohol use and facial flushing to blood pressure and hba1c among Cambodian populations with dysglycemia in the U.S. and in Cambodia. Diabetes & Metabolic Syndrome: Clinical Research & Reviews, 102374. https://doi.org/10.1016/j.dsx.2021.102374

Bagdas D, Rupprecht LE, Nunes EJ, Schillinger E, Immanuel JJ, Addy NA. Effects of Cherry and Vanilla Flavorants on Oral Nicotine Liking and Disliking in Rats. In: The Society for Research on Nicotine and Tobacco 26th Annual Meeting; 2020 Mar 11-14; New Orleans, LA. https://cdn.ymaws.com/www.srnt.org/resource/resmgr/conferences/2020_annual_meeting/SRNT20_Abstracts_NEW_0227202.pdf

Bagdas, D., Rupprecht, L. E., Nunes, E. J., Schillinger, E., Immanuel, J. J., & Addy, N. A. (2021). Evaluation of Flavor Effects on Oral Nicotine Liking and/or Disliking Using the Taste Reactivity Test in Rats. Nicotine & Tobacco Research. https://doi.org/10.1093/ntr/ntab241