Mateo Arcos
Experience
Software Developer Intern
Philer.ai
- Developed a voice-based intake agent using Python, LangGraph, LLMs and Twilio; streamed QnA state and output to JSON.
- Built Flask microservice with async routing and REST API for outbound calls.
- Automated Airtable enrichment to prefill forms; reduced call length by 35%.
- Designed and developed User-Interface using HTML, CSS; Deployed containerized service on AWS.
Analytics Engineer Intern
Loblaw Companies Limited
- Created PHP/JS dashboards to streamline Supply Chain reporting (Replenishment Optimization Team).
- Refactored SQL queries; cut dashboard latency by 70%.
- Built ETL pipeline (Python, SQL) for 50k+ daily DC order cuts; quantified impact of cut size on near-expiry pushes via logistic regression, attributing $1.58M in near-expiry inventory to prior cuts.
Finance Risk Analytics Intern
OMERS
- Built a Python script hierarchical clustering to deduplicate 5k+ counterparty names, automating canonical name mapping and eliminating manual CRM data cleanup.
- Engineered ETL ingesting SQL/SharePoint for daily liquidity models.
- Replaced Excel workflows with Centralized data pipelines and model outputs in Power BI; collaborated with cross-functional teams to translate analytics into dashboards for senior stakeholders and CRO.
Research Assistant
AR Environmental Solutions
- Supported literature review on organic food and health outcomes.
- 1st place Sobeys Innovation Case Competition—bioethanol waste-to-fuel pitch.
- Ran statistical analysis on lab tests involving mineral concentrations.
Projects
UofT Course Navigator
PythonReactFastAPIOpenAIFAISS
GitHub- Deployed a full-stack RAG assistant that answers program- and course-related queries across 3k+ courses.
- Python script to scrape timetable HTML into structured JSON and rendered an interactive prerequisite graph for 3K+ courses, streamlining degree-planning decisions.
- Delivered similar course recommendations with a hybrid semantic + node2vec embedding pipeline.
LLM Multi Agent Task Automation
PythonLangGraphMulti-AgentAutomation
- Developed multi-agent LLM system for automated task orchestration and workflow management.
- Implemented LangGraph framework for complex agent coordination and state management.
- Presented research findings at CUCAI 2025 conference on AI automation techniques.
MedACL – MRI Analyzer
PythonKerasPyTorchGCP Vertex AI
GitHub- Deployed and trained U-Net model for MRI ACL-tear segmentation + CNN classification model on GCP Vertex AI.
- Wrapped in a FastAPI container, and served through Vertex AI Endpoints for real-time inference.
Research Paper Recommender
PythonPyTorch
GitHub- Built Denoising Adversarial Autoencoder to embed research paper titles.
- Simulated noise patterns; optimized via L2 loss.
NFL Third Down Success Rate
RShinyTidyverse
GitHub- Generalized Linear Mixed Model (GLMM) to analyze third-down football plays, catering to the hierarchical data structure from games to teams.
- Trained Regression Model
- The final model adequacy was ascertained by plotting deviance residuals against predictors and calculating the AUC of the ROC curve.
Note Net
PythonLLMsReact.js
GitHub- Convert lecture recordings and handwritten notes into a personalized knowledge graph. It extracts key topics and concepts from audio and image inputs using AI to help students.
