/* I blend technical expertise with a focus on impactful data solutions, driven to enhance my skills and career in AI and data science. */
🟢 Open to work
Currently an Extern @PfizerExplore my portfolio showcasing diverse projects in machine learning, data visualization, and predictive analytics that drive real-world impact.
Created a hybrid stacked model with Scikit-learn and XGBoost to predict 3 factors of long-term financial stability and presented actionable strategies to improve financial stability, specifically lowering debt.
Developed a predictive model with Python to forecast TTC subway delays to present 3 actionable recommendations to several stakeholders on strategies to improve handling TTC delays.
Developed a stock price prediction linear regression model to derive 6 trading and hedging strategies. Ranked top 5 out of several teams for effective hedging strategies and predictions.
.jpg)
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
I blend technical expertise with a focus on impactful data solutions, driven to enhance my skills and career in AI and data science.
I'm Julia Tan, a junior pursuing a Bachelor's in Actuarial Science at the University of Toronto. I served as a Technical/Operations Lead at Riipen Labs, with experience as an Artificial Intelligence Researcher at Algoverse. I'm passionate about harnessing data to drive decisions and improve systems.
Externships
Pfizer Advanced: AI-Powered Document Insights & Data Extraction Externship
Pfizer
Experience
Technical/Operations Lead
Riipen Labs · Mar 2026 - Apr 2026
Artificial Intelligence Researcher
Algoverse · Jun 2025 - Jan 2026
Education
University of Toronto
B.Sc. Actuarial Science and Economics · Class of 2028
Skills
SDSS Datathon · Mar 2026
Created a hybrid stacked model with Scikit-learn and XGBoost to predict 3 factors of long-term financial stability and presented actionable strategies to improve financial stability, specifically lowering debt.
Created a hybrid stacked model with Scikit-learn and XGBoost to predict 3 factors of long-term financial stability and presented actionable strategies to improve financial stability, specifically lowering debt.
SDSS Datathon · Feb 2025
Developed a predictive model with Python to forecast TTC subway delays to present 3 actionable recommendations to several stakeholders on strategies to improve handling TTC delays.
Developed a predictive model with Python to forecast TTC subway delays to present 3 actionable recommendations to several stakeholders on strategies to improve handling TTC delays.
Quantitative Finance Club · Jan 2025
Developed a stock price prediction linear regression model to derive 6 trading and hedging strategies. Ranked top 5 out of several teams for effective hedging strategies and predictions.
Developed a stock price prediction linear regression model to derive 6 trading and hedging strategies. Ranked top 5 out of several teams for effective hedging strategies and predictions.
✅ Verified by Extern · ⏱️ In progress
Prototype AI-powered document intelligence with Pfizer—using OCR, LLMs, and RAG to automate real enterprise PDF workflows and build a standout portfolio project.
The externship prototyped AI-powered document intelligence workflows for enterprise PDFs. The work combined OCR, large language models, and retrieval-augmented generation to process and extract information from documents. Deliverables included research, architectural notes, and proof-of-concept extraction pipelines.
I produced a technical summary that explained tokenization, embedding formation, next-token prediction mechanics, weight updates during training, and how models select the highest-probability output token.
The submission summarized how LLMs tokenized input, converted tokens to numeric embeddings, learned word relationships by next-token prediction during training, adjusted weights from errors, and produced the highest-probability token as output.
During the externship I loaded raw document files in Colab, cleaned structured and unstructured text with Pandas and regex, processed JSON into normalized tables, and applied OpenCV denoising and histogram equalization to enhance a scanned document image.
The project processed SDF PDFs to extract text and bounding boxes, identified where tables and special symbols were missed, and noted that table structure and date formatting were not preserved. The deliverable summarized these extraction results.