Data Engineer

TODOR ANTONIJEVIC, PhD

PROFESSIONAL SUMMARY

Enthusiastic and self-motivated data scientist with strong analytical skills and the ability to function in interdisciplinary research settings as demonstrated by development and deployment of a new machine learning algorithm that infers underlying network interactions from complex high-throughput in vitro time-course data.

Result-oriented scientist with 6 years of experience in Python, statistical, analytical and mathematical modeling as evidenced by building new predictive models to evaluate the chemical safety and improve risk assessment.

Strong Python and R skills including: parallel processing, writing high performance algorithms (python code is just-in-time compiled to increase speed up to 50x), machine learning, and server execution.

WORK EXPERIENCE

POSTDOCTORAL FELLOW EPA National Center for Computational Toxicology / July 2015 – Present

Development and application of new computational models to evaluate chemical safety. This research is part of a new vision for toxicity testing in the 21st century where predictive computational models based on high-throughput screening will replace current animal studies. Specific tasks:

Independently developed a new machine learning algorithm to infer network responses from multidimensional time-course data following chemical treatments. Carried out all steps of large-scale data processing including: normalization, statistical analysis, data visualization, and data mining. Math-heavy Python code was just-in-time compiled and executed on servers using parallel processing.

Improved and implemented a new predictive model to forecast a critical transition in system’s behavior (“tipping point”) between adaptation and adverse effects after a chemical treatment. Extrapolation of critical concentrations produced oral equivalent doses lower than animal sub-chronic studies.

Successfully predicted chemical toxicity in animals by extrapolating experimental treatments to oral equivalent doses and performing machine learning (using the scikit-learn library).

GRADUATE RESEARCH (Ph.D.) the University of North Carolina at Greensboro / July 2011 – May 2015

Supported development of a novel, noninvasive method to assess risk for heart conduction instabilities by fitting Chernyak-Starobin-Cohen (CSC) model to ECG data.

Wrote an algorithm to computationally model critical phenomena in biological systems by implementing statistical Metropolis Monte Carlo Simulations.

PROJECT ENGINEER Konkav Konvex, Serbia / Nov. 2006 – July 2009

Performed feasibility study, and time management for each project.

Designed nonstandard parts needed for successful project finalization, and coordinated delivery with suppliers.

Prepared working orders for production and supervised production

EDUCATION

The University of North Carolina at Greensboro

Doctor of Philosophy (Ph.D.) in Nanosciences May 2015

North Carolina Central University, Durham, NC

Master of Sciences (M.Sc.) in Physics Dec. 2011

The University of Belgrade, Belgrade, Serbia

Bachelor of Science (B.Sc.) in Mechanical Engineering Oct. 2006

TECHNICAL EXPERIENCE

PYTHON

R

DATA MINING

NumPy

SciPy

Scikit-learn

Pandas

Matplotlib 2D and 3D

MACHINE LEARNING

PREDICTIVE ANALYTICS

UNIX / LINUX / WINDOWS

Supervised learning

PARALLEL PROGRAMMING

Matlab / Octave

Unsupervised learning

NUMERICAL MODELING

Web scraping

Statistics

Interaction with:

mySQL

mongoDB - NoSQL

Contact this candidate