CV
Basics
Name | Kaloyan Parvanov |
Label | Machine Learning Specialist |
parvanovkaloyan@gmail.com | |
Phone | 781-346-5802 |
Url | https://www.kparvanov.com |
Summary | Machine Learning specialist with a strong background in Applied Mathematics and a proven ability to enhance AI model performance through meticulous data annotation. Consistently received top evaluator ratings (average 4.8/5) for quality and accuracy. Proficient in Python and ML frameworks; eager to drive innovation in AI and Data Science projects. |
Work
-
2024.07 - Present AI Training Specialist
DataAnnotation & Outlier AI
Freelance position enhancing AI models and collaborating with development teams.
- Enhanced AI models by correcting coding and mathematical responses, improving accuracy and reliability.
- Identified and rectified hallucinations in AI outputs, contributing to significant error rate reduction.
- Consistently achieved high evaluator ratings (5/5 excellent 65% of the time) for quality and precision of annotations.
- Collaborated with AI development teams to refine algorithms, leading to more precise AI solutions.
-
2021.08 - 2024.05 Graduate Teaching Assistant
University of Colorado Boulder
Assisted in teaching Calculus and Differential Equations courses.
- Assisted in teaching Calculus and Differential Equations courses to over 100 students.
- Developed and graded assignments, providing constructive feedback that enhanced student understanding and boosted exam performance.
Education
Skills
Programming Languages | |
Python | |
R | |
SQL | |
C++ | |
JavaScript |
Machine Learning Frameworks | |
TensorFlow | |
PyTorch | |
scikit-learn |
Deep Learning | |
Neural Networks | |
NLP | |
Physics-Informed Neural Networks (PINNs) |
Data Analysis | |
pandas | |
NumPy | |
SciPy |
Data Visualization | |
Matplotlib | |
Seaborn | |
Power BI |
Web Development | |
FastAPI | |
Next.js |
Tools & Platforms | |
Git | |
Docker | |
LaTeX |
Languages
English | |
Fluent |
German | |
Fluent |
Bulgarian | |
Fluent |
Projects
- 2024.08 - 2024.10
Regime Shift Detection in S&P 500 Stocks
Applied PCA and Sparse PCA to analyze regime shifts in S&P 500 stocks using high-frequency data.
- Implemented full period and rolling window PCA analysis on 2-minute interval data for 457 S&P 500 stocks over 42 trading days.
- Developed interactive visualizations to showcase sector-based PCA, correlation with ETF factors, and intraday pattern analysis.
- Identified key market drivers and potential regime shifts, particularly around the September 2024 Fed rate cut.
- 2024.08 - 2024.09
MathBuddy: AI-Powered Math Tutor
Full-stack AI tutor utilizing GPT-4 for interactions and GPT-3.5-Turbo for result extraction and difficulty estimation.
- Engineered a full-stack AI tutor utilized by over 200 users.
- Implemented serverless architecture with Next.js frontend and FastAPI backend, integrating OpenAI and Wolfram Alpha APIs to enhance problem-solving capabilities.
- 2024.06 - 2024.06
Tic-Tac-Toe with Alpha-Beta Pruning
Tic-Tac-Toe game featuring an AI opponent using the Alpha-Beta Pruning Minimax algorithm.
- Developed a Tic-Tac-Toe game featuring an AI opponent using the Alpha-Beta Pruning Minimax algorithm.
- Improved AI decision-making speed by 40% by reducing evaluated nodes, enhancing gameplay experience.
- 2023.10 - 2023.12
ODE Solution via PINNs
Solved the damped unforced pendulum problem using Physics-Informed Neural Networks.
- Solved the damped unforced pendulum problem using Physics-Informed Neural Networks, demonstrating effectiveness in complex ODEs.
- Achieved 15% higher accuracy compared to traditional numerical methods, validating the potential of PINNs in solving differential equations.