Peep Küngas

School of Exact Sciences at University of Tartu

Norwegian University of Science and Technology (NTNU)

TalTech University

★★★★★ Data and Services Practitioner

Tallinn, Harjumaa, Estonia

Education and Youth Board of Estonia

Data Architect

Raising capital to finance business, which will challenge the global business credit scoring market. Refining product concept and maintaining product development roadmap. Developing credit scoring models by using machine learning and artificial intelligence. Developing digital marketing strategy and executing digital marketing activities for user acquisition. Having fun with a great team of professionals:)

Scoriff

Founder

Developing Data Quality Framework, which will be part of Estonian national Data Governance Framework. The main emphasis is set to applicability in public sector organizations.Consulted Estonian public sector with Big Four consulting firms PwC and EY in areas of semantic interoperability, linked data, IT strategies, e-commerce and e-business. Compiled and led execution of national strategies for Web and semantic interoperability in Estonia. Conducted ontology engineering and semantic annotation workshops for specialists.Developed ontologies and data schema for 20+ open and closed datasets, including taxes, finances, debts, organizations, board membership and ownership, buildings, real estate ownership, addresses, official notices, Web mentions etc, in Estonia. Designed Graph API endpoints for exposing open data as linked data via API-s.Assembled and coordinated an agile team, which developed, using LEMP stack, inforegister.ee Web site from a rough concept to a sellable product, which generates today 1MEUR+ annual revenue. Applied various growth hacking tactics, including SEO, microdata, SEF URLs etc, to increase monthly visits of the Web site from 0 to ca 3M, where ca 90% of traffic is organic and the rest mostly direct. Maintained the solution until its deployment at cloud. Applied log mining and simpler analysis techniques to identify DDoS attack patterns and required alterations to Web server configuration to effectively mitigate the attacks.Applied graph search algorithms on graphs with millions of nodes and tens of millions of edges for fraud analytics and investigation of corruption cases.Developed and operated a public online Web services catalogue <a href="http://soatrader.com">soatrader.com</a> which brokered access to tens of thousands global Web services.Reviewed and evaluated research and development grant applications in the area IT wrt economic impact, innovation and other aspects.

SOA Trader

Founder and Owner

Carrying out daily data operations in 3 core threads: data governance, data analytics and data science. In data governance thread I am standardizing and documenting data models in the field of education by using linked data principles. In data analytics I am making sure that real-time click-stream data from dozens of Web sites, combined with less-frequently changing data, is available for operational decision-making via mainstream analytics tools and dashboard systems. Finally, in the data science stream I am responsible for prototyping, specification and development of personalized learning path infrastructure, which simplifies harvesting, ingestion, labeling and usage, including advanced learning analytics, of behavioral data of learners' in EdTech solutions nationwide.

HITSA - Estonian Information Technology Foundation for Education

Built and coordinated a top-grade data science team of 6 data scientists and engineers in projects involving Web mining, text mining, log mining, text analytics, clustering, supervised machine learning, Hadoop, Spark, HDFS, R, Python, Java, ...Coordinated product development and R&D activities with infrastructure, software development and product partners. Aligned in-house digital marketing team activities with product development activities and ensured digital marketing support in software products.Raised EUR 672,000 funding for a joint credit scoring project for years 2017-2019. Raised EUR 400,000 funding for a joint process mining project for years 2015-2019.

Register OÜ

Development Manager

Managed a software product portfolio consisting of a data product with ca 20 features, business credit risk prediction product, B2B lead scoring product for telemarketing use, a mobile credit management product and a desktop product for due diligence. Maintained product backlog, collected and annotated user feedback, wrote user stories and use cases, elicitated API and end-user product requirements. Worked with development team to ensure agile implementation of product features, improvements and bug fixes.Implemented business credit risk prediction product starting from idea, requirement elicitation and business cases, followed by implementation of data processing pipelines in R and ending with adoption in end-user products and implementation of marketing/sales collateral. Used machine learning methods and ensemble learning to train a business credit risk prediction model with accuracy of 99.8% and implemented batch processes for daily credit scoring for all businesses in the Estonian market.

Product owner

Developed algorithms to annotate semi-automatically large sets of XML and database schema. Designed ontologies for annotation of interfaces of information systems. Combined text mining and ontology learning algorithms with social network analysis to reveal interoperability issues in large information systems by using their interface descriptions. Designed and applied algorithms, using social network analysis and other metrics, for analyzing information diffusion, cohesion, redundancy and other phenomena in state-scale information systems. Developed means to visualize state-scale services networks and data redundancy.Reverse-engineered reputation of scientific conferences by using decision tree learning methods. Used process mining tools and log mining techniques to reveal state service usage patterns from X-Road service invocation logs.Worked as lecturer for Enterprise Software seminar and Ontology Engineering course. Supervised ca 40 student thesis projects at bachelor (12), master (24) and doctoral levels (3). Participated FP7 projects include LiquidPub, ACSI, REMICS, GET SERVICE, and Horizon 2020 project SoBigData. Raised funding for both national and international research projects.Conducted research and published scientific articles in Web services, automated semantic annotation, ontology learning, Web application synthesis, linked data, (social) network analysis, national IT infrastructure, information diffusion, scientific reputation, semantic interoperability, big data,  ...

University of Tartu

Senior researcher

Managed applied research projects and supervised teams in 1) content extraction, analysis and aggregation, 2) enhanced usability of rich Internet applications, 3) cross-browser compliant rich Internet application design & testing), and 4) agile interoperation in service networks. Developed and maintained a portfolio of business cases and matching experiments.Applied hierarchical clustering and other machine learning methods for optimizing automated cross-browser testing and for classifying cross-browser compatibility issues. Used named entity recognition and other NLP and text analytics methods for feature extraction from unstructured text content. Applied sentiment analysis of online comments by using robust approach with Naive Bayes and a bag of words.Specified requirements and use cases for a configurable SOAP proxy cache for data provisioning Web services. Devised and executed rigorous performance testing, which showed that with SOAP request normalization (C14N, N10N) and LRU cache replacement policy delay savings ratio up to 80% and byte hit ratio over 80% can be achieved in practice.

Software Technology and Applications Competence Centre

Post-doctoral position at Centre of Excellence in Computer Science, EXCS

Researcher

Designed and implemented (in Java) Web services composition / selection, invocation and knowledge aggregation middleware for robot swarms by using AI planning, knowledge management, ontologies and other technologies.

Royal Institute of Technology

Post-doctoral fellow

Implemented in Java the first domain-independent resource-conscious AI planner, which used linear logic theorem prover as core engine, and demonstrated performance similar to state-of-the-art domain-independent planners on standard AI planning benchmark datasets. Implemented distributed AI planning algorithms in Java for supporting collaborative problem solving and distributed resource allocation in multi-agent systems.Implemented a Web services indexing and search engine, which harvested WSDL/XSD descriptions of Web services from the Web and UDDI directories, annotated XML interfaces semantically and provided automated web services composition as search on top of the semantic descriptions. Used the search engine to analyze variety and availability of data at global scale. Also used the dataset also to identify automatically potential integration points of Estonian governmental X-Road Web services and global Web services.Worked as teaching assistant for Logic course, lecturer for Logic and Planning course and teaching assistant for Distributed Artificial Intelligence and Intelligent Agents course.Conducted research and published scientific articles in multi-agent systems, (distributed) AI planning, linear logic, partial deduction, cooperative problem solving, Web services, semantic Web services composition, distributed AI, intelligent agents, P2P computing, search algorithms, ...

Norwegian University of Science and Technology

Research Fellow

Implemented Chord P2P protocol and algorithm in Mozart (<a href="http://mozart.github.io/">http://mozart.github.io/</a>) programming system and run simulations, which revealed a major flaw in the algorithm. Proposed advancements to Chord to fix the flaw and validated the fix via simulations.Developed a search space reduction technique for Petri net reachability checking by using abstraction hierarchies - the technique allowed to effectively tackle state explosion problem under certain conditions.

Visiting Researcher

Elicited requirements from field robotics in the context of AI planning. Advanced an AI planning system wrt requirements of field robotics. Adopted Petri net reachability checking heuristics to dramatically increase performance of a resource-conscious AI planning algorithm via reduction of impact of state explosion problem.

Helsinki University of Technology

Implemented machine learning and search algorithms in C and Java for supporting automated software synthesis and robot navigation. Implemented a general-purpose resource-conscious AI planner engine in Java. Worked as teaching assistant for Theoretical Computer Science course at TalTech. Conducted research and published scientific articles in AI planning, linear logic, automata learning, Petri Net reachability checking, software synthesis, ...