Looking for a job

Canyang Zhao

@canyangzhao

Data science and software development intern at Bio-Techne who built an Azure research analytics app that cut analysis turnaround time by a third.

United States

Message

What I'm looking for

I’m seeking a Summer 2027 internship in data science, analytics, or data engineering.

At Bio-Techne, I built an Azure-hosted Python API and TypeScript application for scientists, replacing commercial analysis software with configurable filtering on internal research data. It saved approximately $10K monthly and cut analysis turnaround time by a third.

I also trained an XGBoost ranker for R&D decision support, improving top-1 selection success from a 50% baseline to 70% on held-out runs. I supported a 200 TB migration to ADLS Gen2 by refactoring data access and validating transfer integrity.

At AI Rudder, I fine-tuned Llama 3.1 8B with LoRA SFT on customer-service tool-call traces. Fine-tuning raised strict tool-call correctness to 80% on the held-out regression suite, and I helped build release evaluations that routed disagreements for human review.

At Turning Green, I built an Airbyte-to-Snowflake ELT pipeline and SQL models for site-selection reporting, reducing manual data preparation from four hours to 30 minutes per update. I’m pursuing an M.S. in Data Science Biomed at the University of Wisconsin–Madison and am building an autonomous coding agent as a machine learning hack project.

Experience

Work history, roles, and key accomplishments

Graduate Intern, Software Development

Bio-Techne

May 2026 - Aug 2026 (3 months)

Built an Azure-hosted Python API and TypeScript application for 50+ scientists, replacing commercial analysis software and saving approximately $10K monthly. Developed an asynchronous Celery service with PostgreSQL and Docker, sustaining 50+ completed jobs per workday at 99% completion and cutting median queue latency from 12 to 4 minutes.

Python TypeScript Azure PostgreSQL Docker Celery ADLS Gen2 XGBoost Databricks ML Inference

LLM Engineer Intern

AI Rudder

May 2025 - Aug 2025 (3 months)

Fine-tuned Llama 3.1 8B with LoRA SFT on customer-service tool-call traces, reducing incorrect or missed calls from 12 to 9 per 100 eligible turns. Built release evaluations combining tool-call checks with a three-level LLM judge, achieving 79% agreement with human judgments.

LLaMA LoRAs SFT LLM Evaluation Tool calling Python Prompt Engineering

Data Analyst Intern (Data Engineering)

Turning Green

Dec 2024 - Mar 2025 (3 months)

Built Airbyte-to-Snowflake ELT for 18 months of CDFA, school, and market data, reducing manual data preparation from four hours to 30 minutes per update. Created a Snowflake feature pipeline for proximity analysis and trained an XGBoost viability model to rank hub sites.

Airbyte Snowflake SQL XGBoost Data Engineering ELT Feature Engineering