Guntas Toor

About

Hi, I’m Guntas. I'm a Computer Engineering student at Queen’s University in Kingston. I’m in a five-year program that includes a PEY term, and I expect to graduate in 2027/2028 depending on Co-Op terms.

I’m currently a Software Engineering Intern at Sentry in San Francisco, and I do ML research on the side. Technically, I’m most interested in software development, systems-level infrastructure, artificial intelligence, and quantitative finance.

Beyond academics, I enjoy traveling, going to the gym, exploring new foods, and playing chess. I use this website to document my learning and lifestyle and showcase projects as I grow as an engineer.

Skills

Experience

• 
  May 2026 — Present San Francisco, CA Now

  Software Engineering Intern at Sentry.io

  • Built a LangGraph agent that traces ARR drops through our dbt + Airflow lineage and points at the likely cause — took diagnosis from a few hours down to minutes (~15x faster).
  • Wired up a vector search across ~12M tokens of Looker dashboards and Git commits so it can actually find the relevant context, landing around 91% retrieval accuracy.
  • Deployed the whole thing on GCP with FastAPI + Redis; it quietly clears ~15% of tickets on its own and saves the team 7+ hours a week.
  • Shipped a few-shot LLM plus a distilled classifier to tag 600k customers by industry at ~95% F1.
• 
  September 2025 — Dec 2025 Remote

  Applied ML Researcher at Algoverse

  • First author on a mechanistic interpretability paper, accepted at the ICLR 2026 LMRL Workshop (paper link).
  • Built an activation patching pipeline over 32-layer LLMs to figure out where pharmacological knowledge actually lives across the MLP outputs.
  • Trained linear probes on the model activations and got them to ~0.86 ROC-AUC classifying medical drugs.
  • Put together a 6K+ sample benchmark with automated RAG retrieval, which bumped reasoning evaluation by 30–40%.
• 
  May 2025 — April 2026 Toronto, ON

  Software Engineering Intern at Flipp

  • Wrote Scala microservices that scale the flyer ETL pipelines over terabytes of data in S3, serving 10M+ users.
  • Moved our Python services off Amazon MWAA onto EKS with Docker, which trimmed Airflow costs by ~20%.
  • Kept the real-time streaming systems sitting at 99%+ uptime (tracked on Datadog) using Kafka and Spark Streaming.
  • Automated migrating 1000+ schemas into Databricks with multithreaded SparkSQL, cutting the runtime by ~60%.
  • Wrote up some Cursor + Claude AI workflows for the team, which sped PR reviews up by ~20%.
• 
  May 2024 — Dec 2025 Remote

  Software Developer (Part Time) at Optimize Everything

  • Redesigned the PostgreSQL schema and tuned the queries, which got performance up around 40%.
  • Built an auction system for a mobile transit platform that picked up 2000+ beta users, working in an 8-person agile team.
  • Wired up 5 REST APIs with SQLAlchemy to keep the frontend and backend talking properly.
• 
  Sep 2023 — Apr 2025 Kingston, ON

  Machine Learning Researcher at Queen's University AI Institute

  • Tuned an LSTM for stock price forecasting with a bunch of time-series preprocessing and got it to ~95% accuracy.
  • Built an NLP pipeline over 1500+ financial articles to see how the news actually correlates with stock prices.
  • Shipped a full-stack Flask app that gives an AI read on stocks, and presented it to 320+ delegates at CUCAI.
• 
  Sep 2023 — Dec 2025 Kingston, ON

  Embedded Systems Engineer at Queen's University Hyperloop Design Team

  • Looked after the thermal sensors for the electronics packed into the hyperloop chassis, keeping tabs on the readings.
  • Wrote the sensor logic in C++ on a Raspberry Pi and tuned it to keep everything running within temperature.

Projects

• ShopPay 2025 · Python · SQL · AWS

  Collaborated on a hackathon web app that lets users input Amazon product links and recommends similar items from local businesses. Leveraged Amazon Comprehend for text analysis, scikit-learn for similarity, and SQL for product storage.

• Movement Classification App 2025 · Python · Scikit-learn · NumPy

  Created a real-time movement classification system that distinguishes walking vs. jumping with ~98% accuracy on 300k+ samples using a Logistic Regression model. Delivered both a desktop client (Tkinter + Matplotlib) and a web app (FastAPI backend with Tailwind + Chart.js) with live PhyPhox sensor integration via Selenium for real-time streaming and classification.

• Security System 2025 · C++ · NodeJS · HTML/CSS

  Developed an RFID-based door access system with embedded C++ for tag detection and access control, paired with a Node.js backend and web dashboard. Designed the system with a clear separation between hardware logic, API services, and user interface for reliable and extensible access management.

• COVID-19 Mutation Region Classifier 2025 · Python · Matplotlib · Seaborn

  Built a machine learning model to predict the geographical origin of COVID-19 mutations using logistic regression. Implemented genomic feature engineering, data balancing, and model evaluation to achieve strong classification accuracy across Asia, North America, and Oceania.

• Social Media Platform 2024 · C · Data Structures & Algorithms

  Designed and implemented a social media platform from scratch using C. Developed efficient data structures for user profiles, friendships, and posts, plus messaging and feed systems. Explored trade-offs of linked lists, trees, hash tables, and queues to ensure scalable performance.

Travel & Food

Contact

• Email: guntastoor@gmail.com
• GitHub: github.com/Guntas07
• LinkedIn: linkedin.com/in/guntastoor