It is time for our next Meetup. In March, Capgemini will host for the first time.
• 2 presentations (each ca. 30-40 min incl. discussion)
• Of course time for networking + food + drinks before, in between and especially after the presentations
• Talks are held in English
The line up:
Daniel Kühlwein - Saving the environment with DeepLearning
Oil is arguably one of the most important resources of our modern world. Unfortunately, it can also do great harm to our environment:
Bursting oil pipelines can destroy complete ecosystems. How can we make sure that this doesn’t happen?
One aspect is pipeline maintenance. This is currently done by robots that travel through the pipeline and scan the inside.
The resulting images are then used to detect potential defects. As data scientist, this naturally begs the question: Can we use deep learning for this?
The short answer is ‘yes’, and it works amazingly well.
The presentation will give the long answer: the data we had to deal with, the models we tried, what worked and what didn’t.
Daniel Kühlwein leads the Data Science Community at Capgemini and is still in awe of the power of deep learning. Before joining Capgemini, he taught automated theorem provers mathematical intuition, build a proof checker for (almost) natural language mathematical proofs and interned at MS research.
Ankit Bahuguna - Deep Learning: Applications in Information Retrieval
A web search engine allows a user to type few words of query and in return presents a list of potential relevant web pages within fraction of a second. Traditionally, keywords in the user query were fuzzy-matched in realtime with the keywords within different pages of the index and they didn’t really focus on understanding meaning of query. Recently, Deep Learning + NLP techniques try to represent sentences or documents as fixed dimensional vectors in high dimensional space. These special vectors inherit semantics of the document.
The talk will cover the search system at Cliqz from the perspective of applications of Deep Learning. Some of the sub-topic covered are:
- Query Embeddings: An unsupervised deep learning based system, which recognizes similarity between queries through their high dimensional vectors.
- Approximate Nearest Neighbor Systems (ANN)
- Index Freshness and Search Recency - Kubernetes in Production
- Recent work, explorations and enhancements.
Ankit Bahuguna is a Software Engineer (R & D) working in the core web search team at Cliqz GmbH, where he is responsible for improving both Search Quality and Ranking of the search engine powering the Cliqz browser which is currently available in Germany, US and France. Ankit has a master's degree in Computer Science from Technical University of Munich, with research experience at LMU (Germany), IIT Bombay (India) in areas comprising Deep Learning, Machine Learning and Natural Language Processing. He has been a volunteer open source evangelist since many years and currently officially represents Mozilla through its Reps program.
Ankit Bahuguna: https://linkedin.com/in/ankitbahuguna/