Peter Tegelaar

Leading and growing Klarna’s data science organisation of 50+ data scientists. Also: - Jan2020 - now: part of the support team of the largest product domain in the company (230+ engineers, data scientists, product managers, analysts, etc.). - Jan2019 - Dec2019: part of the support team of the largest engineering domain in the company (170+ people). Acting domain lead in domain lead's absence.

Continued to lead and grow the data science organisation I set up, and now reporting directly into the CEO. My team created recommendation engines and algorithms for image classification, pricing, automatic curation of lots, fraud detection, expert augmentation & workflow automation, and many more use cases.

Setting up the data science, analyst, and BI/DWH organisations in 2 offices. Reported into the Founder CTO.

Set the research directions in algorithmic mobile advertising, did product management, and recruited & led an international team of 4 Ph.D. data scientists. The team improved the existing system of algorithms and bidding strategies to such an extent, that the average cost-per-click (CPC) decreased by 50%, significantly affecting the bottom line of the company. In addition, I continued my day-to-day data science work. NB LiquidM Inc. is a Madvertise technology spin-off.

Researched, conceptualised and designed the decision-making & bidding logic of one of the 1st mobile-only RTB advertising DSPs (demand-side platform) in Europe. A DSP brings together machine learning, reinforcement learning, time series analysis, and bidding strategies into a coherent system. Worked closely together w/ data engineers to turn my research & prototypes (in Python and R, with data pipelines built in Apache Pig) into production code, including direct commits of my own algorithms (in Haskell).

Stress test analysis and building tools for credit risk exposures in emerging markets.

Running stress tests and building tools for credit risk exposures in emerging markets. Also performed rating analyses of SPVs with bond underliers, and of Dutch midcaps.

We were the first Dutch company to be funded by Y Combinator, the #1 seed capital fund in the world (2% acceptance rate). The team spent 3 months in Silicon Valley, for which we were featured in Businessweek: http://bit.ly/96S1aA. Primarily responsible for product & algorithmic development. Also did recruiting & project management (team up to 6 people), business development (secured pilot with a top 10 game on Facebook with 4.5M monthly active users), and fund raising.

Created “Ludwig”, a production-ready collaborative filtering recommendation engine, which recommended new software and games to users of the Wakoopa tracking software. The code was written without the help of any machine learning libraries - which were unavailable at the time - in Ruby with C extensions for performance. A month after finishing at Wakoopa, I competed with my own startup project Reccoon in the 2008 'Strands $100,000 Call for Recommender Startups' competition, where we came in second place.

Research into equities, futures, and commodities. Analysed statistical arbitrages and did some modeling. Also learned to work with tools ubiquitously used in the financial industry (Bloomberg, Excel, etc.)

Propedeuse (=1st year exam): Computer Science (2007) Minor: Physics (2009) Thesis (9.0/10) - in preprint: "Combining Kernel Estimators in the Bivariate Uniform Deconvolution Model." http://arxiv.org/pdf/1101.0935v1

Focus on Microeconomics, Behavioral Economics, and Industrial Organisation. Thesis (Grade 9.5/10): "Overconfidence and Behaviour in Economic Games" http://dare.uva.nl/document/41898

Specialisation: International Financial Economics International program: 80% foreign students, all courses in English Thesis (8.5/10): "Optimal venture capital financing of high-tech start-ups. Minimizing agency costs in a dynamic model"

Spent a semester studying abroad in Lisbon (PT) in 2005.