Banu Selin Tosun

Currently: Sr. Data Scientist at Neal Analytics (FTE at Neal & working with Lenovo)
Contact me: selin.tosun@gmail.com
Professional Profiles:
Profile
- Data scientist with a detailed understanding of mathematics and applied statistics in a Python environment, with 10+ years of experience in analytical problem solving
- Hands-on experience working with big data (Apache Spark) with a focus on finding meaningful conclusions
- Experience in experiment design
- Experience in explaining and presenting results in context
Technical Skills
- Data science & analysis expertise using Python (pandas, NumPy, Seaborn, Scikit-learn, TensorFlow, Caffe2), Spark, SQL, MatLab,
and well-versed in Machine Learning, Feature Engineering, Experimental Design, Statistical Analysis
- Git, Hadoop, AWS, Azure, QGIS, Horovod, Docker, Octave, AWS, MS Azure, CNTK and basic knowledge of HTML, JavaScript, & CSS
- Material characterization and processing expertise in both research and manufacturing environments
Data Science Projects
- Preventative Maintanance with Lenovo LCE Team
- Build end-to-end pipeline for laptop crashes using big data (40 GB daily over the course of 4 months) in Databricks PySpark environment
- Developed Hypothesis driven investigation maps to priorities the value and the collection feasibility of necessary data
- Formed multi-component solution engine that detects Anomalies and predicts (via Clustering & Classification) fail types
- Obtained over 87 % average AUC-PR over the detection of top 10 failure modes
- Object Detection and Behavior Detection with Microsoft Azure Media Services
- 6 months on POC around object detection algorithms and bench marking of various Tensorflow Model Zoo models on performance and computational power along with cost optimization
- Built pipelines to customize trainings and model comparison for end-user in mean Average Precision and Inference time
- Developed a distributed training pipeline by Horovod to shorten the training time by number of nodes
- TakeAPic
- January 2018 GitHub on analyzing the Facial Expressions. I build an in house CNN model to classify the 7 different expressions using a 55K+ image DB. I used an Nvidia GPU along with TensorFlow, Keras, and Cuda process on AWS EC2 instance along with 400 GB attached volumes. The confusion matrix showed over 99% accuracy. The misidentified classes were sadness vs fear and anger vs surprise.
- StreetSmart
- September 2017 GitHub on Real Estate housing prices. For this project, I used AWS EC2 instances to Grid Search the best regressor model with the minimum median absolute percent error. The best result achieved via Gradient Boosting Regressor with 11.3 MAPE.
Selected Patents
- Eray Aydil, David Norris, Ankur Khare, Banu Selin Tosun, and Andrew Wills, “Metal Chalcogenides and Methods of Making and Using the Same,” U.S. Patent, US 61/434854 P
- Eray Aydil, Stephen A. Campbell, Rebekah Feist, and Banu Selin Tosun, “Optoelectronic Devices with Thin Barrier Films with Crystalline Characteristics that are Conformally Coated onto Complex Surfaces to Provide Protection Against Moisture,” U.S. Patent Application Number 61514133
Honors and Awards
- IBM Women in Data Science Scholarship 2017
- 39th Photovoltaic’s Specialist Conference IEEE (Best Poster – Nominated) 2013
- Minnesota Architectural Foundation’s 2013 Thomas F. Ellerbe Scholarship (Finalist) 2013
- Hysitron Travel Grant – Minnesota AVS Chapter (Best Student Poster Award) 2012
- AVS Dorothy M. and Earl S. Hoffman Travel Grant 2012
- University of Minnesota, Doctoral Dissertation Fellowship 2012-2013
- AVS Dorothy M. and Earl S. Hoffman Travel Grant 2011
- University of Minnesota, Ted Davis Scholarship 2008
- Fulbright Scholarship (Declined to accept offer from the University of Minnesota) 2008
- First-Rank Graduation Award from Istanbul Technical University Chem. Eng 2007
- High Honor List through all semesters 2003-2008
- Republic of Turkey Higher Education Foundation Scholarship (90% tuition) 2002-2007
- Ranked among top 0.6% of more than 1,643,000 candidates who took the National University Entrance Test 2002
Selected First Author Publications
- “Enhanced Carrier Lifetimes of MAPbI3 via Vapor Equilibrated Annealing”, J. Phys. Chem. Lett. 6 (13), 2503-2508 (2015)
- “Efficient Continuous-Flow Chemical Bath Deposition of CdS Films as Buffer Layers for Chalcogenide-Based Solar Cells”, IEEE 39th PVSC Conference, 1192-1194 (2013)
- “Cu2ZnSnS4 Nanocrystal Dispersions in Polar Solvents,” Chem. Commun. 4, 3549-3552 (2013)
- “Structure and Composition of Zn(1-x)Cd(x)S Films Synthesized Through Chemical Bath Deposition,” ACS Appl. Mater. Interfaces 4, 3676-3684 (2012)
- “Tin Dioxide as an Alternative Window Layer for Improving the Damp-Heat Stability of Copper Indium Gallium Diselenide Solar Cells,” J. Vac. Sci. Technol. A 30, 04D101 (2012)
- “Improving the Damp-Heat Stability of Copper Indium Gallium Diselenide Solar Cells with a Semicrystalline Tin Dioxide Overlayer,” Sol. Energy Mater. Sol. Cells 101, 270-276 (2012)
- “Sputter Deposition of Semicrystalline Tin Dioxide Films,” Thin Solid Films 520, 2554 – 2561 (2012)