I am a 4th year student studying Statistics & Cognitive Science (Computational Stream) @ University of Toronto interested in Applied ML and Data Science.
📫 kiarash.kianidehkordi@mail.utoronto.ca 🔗 LinkedIn · GitHub
Languages
Python · R · Java · JavaScript · SQL · LaTeX
ML / Data Science
PyTorch · TensorFlow · Scikit-learn · HuggingFace Transformers · Pandas · NumPy · SciPy · Statsmodels · Matplotlib · Seaborn
Tools & Infrastructure
Git · Jupyter · REST · Azure · CUDA · CrewAI
Databases
PostgreSQL · SQL · ChromaDB (vector) · vector databases
Some Certifications I have completed CUDA Accelerated Computing · Machine Learning for Production · Deep Learning
-
🧠 Predicting Human Reading Time with GPT-2 & BERT Embeddings — Preprocessed the 1M+ row Natural Stories dataset and built a sliding-window batching pipeline to parallelize GPT-2 & BERT (~110M params) forward passes on CUDA. Analyzed surprisal scores and fit linear mixed-effects models to predict reading time — GPT-2 outperformed BERT (AIC improvement of 571 points, p < 1e-100).
Tech:
Python·PyTorch·HuggingFace Transformers·CUDA·Statsmodels·Pandas·NumPy -
⚕️ Parkinson's Classification & Symptom Profiling from Accelerometer Data — Engineered tremor-specific features from wrist-worn accelerometer data (400+ participants) using signal processing — bandpass filtering (3–12 Hz) and power spectral density. Applied Gaussian Mixture Model clustering to profile symptoms without labels (silhouette score 0.7), and validated clusters with t-SNE against diagnostic labels.
Tech:
Python·Scikit-learn·SciPy·NumPy·Matplotlib·Seaborn— GMM, t-SNE, signal processing -
📓 Recall — Local Journal with Semantic Search & Weekly AI Analysis — A local desktop journaling app where every entry is embedded into a ChromaDB vector store for semantic search across past entries. Built a RAG pipeline with a CrewAI agent (OpenAI API) that uses the journal as a
search_journaltool to generate weekly reflections and answer freeform questions, designed to reduce hallucinations.Tech:
Python·PostgreSQL·Tkinter·ChromaDB·CrewAI·OpenAI API— RAG -
🎲 Hidden Costs of Sports Gambling — Data Dashboard — An interactive dashboard surfacing the hidden costs of sports gambling for a general, non-technical audience.
Tech:
R·Shiny·ggplot2 -
🌆 Weather Wanderer — Weather-Aware Urban Exploration App — A Java desktop app that integrates the OpenWeather API to deliver weather-aware location recommendations, built with SOLID principles and clean architecture across presentation, domain, and data layers.
Tech:
Java·Swing·OpenWeather API— SOLID, clean architecture
-
Undergraduate Researcher — UofT iSchool, COoKIE Lab (Prof. Anastasia Kuzminykh) · Sep 2025 – present Leading a study to develop a validation rubric for benchmarks that measure Theory of Mind in LLMs, synthesizing psychometric literature into a framework for assessing benchmark tasks.
-
Research Assistant — UofT Computer Science, IAI Lab · May 2026 – present Supporting the development and deployment of a hybrid recommender system combining contextual multi-armed bandits (reinforcement learning) with LLMs for adaptive mental health interventions. Also serving as a double-blind reviewer for a response-adaptive biostatistics literature review.
-
Research Assistant — OISE, Wisdom & Identity Lab · May – Sep 2025 Translated moral-injury subcomponents into a predictive-processing framework in cognitive science and produced publication-quality literature reviews from clinically coded interview data.
University of Toronto — HBSc · Sep 2022 – Dec 2026 Double Major: Statistics and Cognitive Science (Computational Stream) · Minor: Mathematics
Selected coursework: Statistical Machine Learning · Convex Optimization · Regression Analysis · Time Series Analysis & Forecasting · Data Visualization · Software Design · Artificial Intelligence · Linear Algebra II
🥇 Gold Medal, Iranian National Mathematics Competition — selected to represent Iran at the 2019 South Africa International Mathematics Competition (SAIMC)