François Le Lay Email & Phone Number

Setauket- East Setauket, Ny, Us

New York, Us

Work From Anywhere, Remote, Oo

San Francisco Bay Area, California, Us

Kensu is an AI-driven data observability solution that offers companies the opportunity to monitor the health of their data pipelines at the source, therefore short-circuiting the downstream impact of incidents while providing huge productivity gains across the business and data engineering teams.- I led the company’s field engineering efforts for the US and Europe: I focused on our sales enablement strategies on one hand, and leveraged my exposure to our customers’ needs to inform Kensu's integration roadmap on the other hand.- I implemented multiple proof of concepts with growth stage startups, as well as enterprise prospects. - Key integrations with Spark and Databricks, Snowflake, Python, dbt, Matillion, JIRA, Slack, Teams.

New York, New York, Us

What The Farmer's Dog does: - Challenge the behemoths of the dog food industry- All about cooking for our dogs: fresh, human-grade meals, delivered to your door through a subscription.- Turning unconditional love into uncomplicated care!I led a group focused on data infrastructure, BI and data science, as part of a central Data, Strategy and Insights support function. - I supervised the data science work: NLP applied to voice of the customer analysis and CX automation, time series forecasting applied to CX staffing (optimized for customer wait times). - I supervised the efforts needed to improve the reliability of our ETL pipelines by designing failover strategies and data observability mechanisms- I supervised the integration of a newly deployed database architecture in the analytics funnel, as part of a transition of the website from a monolithic application to a set of decoupled services. - I supported the migration to a new CX Saas (called Gladly) into the company's data stack

Us

Hugging Face is leading the charge of open source AI:- by driving adoption of libraries such as transformers, peft, trl, and diffusers, to name a few.- by encouraging the use of the Hugging Face hub to host machine learning models- by offering optimzed inference services, either via an API, or a custom containerIn my short tenure, I am proud to have made a few key hires, supported the GTM motions related to the Optimum container, and established supportive relationships with various contributors across all teams (dotted line).

Stockholm, Stockholm County, Se

Observing the accelerated pace of innovation in NLP, I switched from the managerial track to the IC track, to sharpen my applied science skills. - I advocated for the use of LLMs (Large Language Models) to improve the representation of music entities (embeddings), with the goal to improve a data matching pipeline used to reconcile music tracks and publishing works. - I contrasted this work with strong baselines (naive text matching, LSH) , in order to demonstrate efficiencies. - I also explored probabilistic soft logic (from Lise Getoor's research lab) and the use of 3rd party datasets, to perform knowledge graph identification.- I iterated on a scoring model based on deep neural nets to make use of those improved representations. I did so by incorporating a human feedback loop, thanks to the precious expertise of music experts embedded in the team. - I explored the use of complementary melody vectors provided by the music information retrieval team, to expand and improve the set of candidate entities scored by the system.- I advocated for an end-to-end AI-driven product feature to produce synthetic, personalized audio stories about music, pitched it to leadership team, something that the company was only able to resource later on (post-tenure, now known as AI DJ). The idea was to tie a Natural Language Generation module, a Knowledge Graph traversal algo, and a Text-to-Speech module, all in one pipeline.

Stockholm, Stockholm County, Se

I supported the growth of the company since its early days in the USA, by hiring and managing over 30 engineers and data scientists. My teams have built the mission-critical infrastructure that allows hundreds of engineers to iterate quickly, safely and at scale: - I supported the Analytics tooling squad, building with QlikSense & Tableau, on top of BigQuery- I supported one of the squads responsible for the long term implementation of the company's new experimentation platform (A/B testing on web, mobile, desktop, smart speakers). This new product has now been made available ouside Spotify (july 2023), and it is called "Confidence". - I advocated for the use of Scala, as a go-to language for our data pipelines (scale: petabytes/week, thousands of data pipelines), written with Scio, a home-grown library for Apache BEAM,- I facilitated the use of GPUs for deep learning workloads on GCP.- I supported the alignment of the ML infrastructure with our research and production workflows, using TensorFlow Extended (TFX), and Kubeflow pipelines.- I wrote a collection of tutorial notebooks demonstrating the use of CNNs (convolutional networks) to perform speech identification on audio mel spectrograms, using Tensorflow.- I supported the infrastructure squad responsible for Spotify's data quality tooling and best practices, evangelized through an amazing set of home-grown certification processes.

Paris, Île-De-France, Fr

At some point in history, Viadeo was LinkedIn's biggest competitor in France and other emerging countries. In 2014, the group had a total membership base of over 50 million professionals.I was managing a full spectrum of data-related activities, from business analytics to behavioral CRM. In that context, I built the teams and tools that allowed Viadeo to extract more value from data: I created an engineering team, a machine learning team (including Ph.D. candidates), and a business intelligence team. Our technological focus was on building a unified and scalable data system used for analytics, data quality, ETL management, email (re)targeting and social network recommendations. We were early adopters of the Apache Spark distributed system (Berkeley) and made extensive use of the Scala language. Our data science efforts leveraged a rich open-source ecosystem: R, Python (scikit-learn, theano) or Julia, Hadoop (Cascading, Scalding). I developed my expertise with Amazon Web Services, which we used to test and deploy clusters in a matter of minutes. On the analytics side we were building our own data visualization platform based on state-of-the-art Javascript libraries (D3.js, Sammy.js, Knockout.js) and custom backend technology (Clojure, Amazon EC2).

London, England, Gb

With 24 million registered members across 17 countries, PhotoBox.com is the european leader of the online printing, sharing and storage of digital photos.I was managing the group Business Intelligence team, and reporting to PhotoBox's CTO. We supported finance as well as local european marketing teams (UK, France, Germany, Italy, Sweden...).More precisely I was in charge of the following:- Group Datawarehouse/ETL procedures (Oracle, Talend, Clojure/Cascalog)- Ad-Hoc/Static Reporting- CRM/Emailing platform (Neolane, several millions of emails/month)- Online survey data collection (Web services)- Data mining (segmentation, predictive analytics)- Web analytics (Omniture)Our scope went beyond pure Business Intelligence and reached both CRM and quantitative analysis, which made it very exciting!

Paris, Ile-De-France, Fr

Prior to its acquisition by Rakuten, PriceMinister was the third most popular e-commerce site in France (12 million visits/month, 4+ million members).My missions:- To participate in the deployment of modern corporate reporting infrastructure (Oracle Warehouse Builder, Oracle Database 10g, Business Objects) while developing and maintaining existing reporting systems (Perl, SQL). - CRM project and analysis of business activities through the use of advanced data mining techniques. Skills:- Advanced statistics and data mining (SAS, SPSS, Excel, R open-source framework), emailing, database conception (PowerDesigner), Oracle SQL, PL/SQL, ETL, Perl/shell programming, Linux administration, Apache/Tomcat web server, Business Objects administration, database performance optimization, SQL optimization. - Strong understanding of online marketing principles, online customer behavior and business logic in general. Ability to communicate with sales, marketing, finance or back-office departments using words they understand.

Zurich, Zurich, Ch

Web development for e-business website ALAPAGE.COM (France Telecom). Developed "La Boutique des Salaries", an online website used by France Telecom's staff, using core PHP components initially developed for alapage.

Haarlem, ., Nl

- Realized statistical studies on Lycos Tripod european customer base (over 6 million users), participated in customer scoring to help optimize costs reduction,- Developed Web-based corporate reporting dashboards to be used by european management,- Fully developed the Lycos Partnershop, a marketing platform used to improve sales of the Lycos Webcenter webhosting product line by recruiting reselling partners.

Le Chesnay Cedex, France, Fr

Worked on database conception issues at french National Labs (IRISA Rennes). Subject: "Improving performances of nearest neighbor search through the use of statistical methods". Context: multimedia databases of highly dimensional feature descriptors. C++ coding.

Los Angeles, Us

Worked for "Content Search and Access" R&D team, dealing with computer vision topics. Developed a video player with specific artifical intelligence engine that allowed real-time scene detection (internal use). Used Microsoft Visual C++ (MFC, DirectX). Understanding of specific markovian models used to perform scene detection.

Ms - Statistical Engineering Diploma, Statistics And Information Systems

Ms, Artificial Intelligence

Bs, Applied Mathematics

As, Physics, Mathematics, Computer Sciences, Earth Sciences