Software Engineer - Conversational AI (01/2020 - Now)
Implemented reusable UI components to let users label and review documents, including the "command bar" and the "classification control". Contributed to the "improve model" page as well as the logic and UI used to display conflicts between predictions and labels.
Implemented the e2e flows to let customers create new entities, the entity tree used for labelling and the new details page for ML entities. Implemented a shared control to "announce" result of actions to users with disabilities.
Contributed to making Bots easier to build, by adding an Intellisense capability to Bot Framework Composer's fields: Implemented a reusable hook and input wrapper component communicating through LSP to an Intellisense server.
Program manager - Windows feedback (05/2016 - 12/2019)
Designed a context-customized feedback form in the Windows Feedback Hub inbox application for improved actionability of feedback based on the specific needs of our key partners, with a current volume of ~600k new feedback entries/month . Lead our first waves of user studies and experiments to improve Feedback Hub end users experience.
Managed our end-to-end contextual survey functionality, including Windows 10 NPS, constantly adding new features to satisfy partner requirements and experimenting with new designs to improve engagement from end users, with a current volume of ~2.3M survey responses/month.
Launched a unified reporting website, enriching feedback with signals from other teams
Defined, built and communicated our KPIs, bringing clarity and empowering the team to be data driven
Developed a cross platform prototype for collecting and analyzing feedback across all Microsoft products with the end goal of delivering the solution "as a service".
Hardware and software\cloud system that can track and safely dispense the correct amount of postoperative prescribed medication, while tracking real-time opioid use and unused medications. Youtube demo.
Assisting pharmacy technicians and pharmacists through use of machine learning and video capture to alert users in real-time of medication dispensing errors in the pharmacy setting. Youtube demo.
Modern equivalent of guidebooks: Outdoors trip planning | Trip reporting | Packing
ActionPacked offers a pre-curated list of adventures, ranging from backcountry skiing to bikepacking. Users can also create their maps with custom routes and waypoints.
ActionPacked can be found on iOS and Android (beta). Maps can be downloaded offline for access in the backcountry.
Users can upload their GPX tracks and share trip reports. Uploaded images can be seen directly on the map, even if taken from a device without GPS.
ActionPacked offers packing tools. Users can easily drag and drop their gear into custom lists. Charts help give insights into pack weight. Lists are cloud based and collaborative.
Home health diagnostic device | Pool of doctors to analyse results online | Platform to cure with sounds
Based on neuro-research, we found that the use of certain sound vibrations can help the body leverage better its natural defenses, helping the brain send the correct "messages" to different organs and thus restore equilibrium.
First step is seeing a medical doctor associated with FeelHarmony to get a prescription. Then, all that is required is connecting to feelharmony.co to access treatments from anywhere, on any device.
Our platform helps customers keep track of their treatment: listening history, self-reported "feeling" over time as well as a way to communicate directly with doctors to dynamically adjust the treatment.
I have conceived, programmed and deployed:
The key to memory is practice. It is very hard to memorize a word or a definition on the very first try. This is where Memloc can help. It takes only one click to save an entry to the cloud from any website with our Chrome plugin. Then you can practice anytime, from the web or your smartphone, and memorize those words for good!
Photo libraries are often a mess: pictures are unclassified and only sorted by date. Therefore, we often get pictures of say work in the middle of memories with our family or friends. Switch address this issue by categorizing pictures before they are even taken. For example, click on the "Work" tab, and every picture will go on the "Work" folder. Of course, you can create and customize your own folders. Everything is built on top of the native photo app for an easy integration.
E-mails can be bucketize in three categories: important information, requested actions and noise. SmartBox leverages NLP and entity extraction to tag e-mails that are deemed important, as well as tag and provide a one click solution to possible actions you might want to take (create a to do - create a calendar event - etc...).
We all want to be healthy and there are tons of apps to help us out there. But we don't feel very accountable when receiving notifications from those apps. We do however feel accountable when we receive messages from our friends. NotifyMe allows you to send friendly reminders to other people or broadcast a message to invite them to join a healthy activity. For example, by just pushing one button, you can send a "Going for a run, who wants to come?" to a group of friends.
A clear view of your day without scrolling. We use colors and emoticons to help you quickly visualize your activities, their type (work, sport, ..) and how long they last. It is also easy to figure out free time when you need to create a new event. You can highly personalize your calendar by creating your own categories, colors and emoticons.
By using SmartDiag, users can input parameters about their house (such as square feet, type of windows) and get in return energetic consumption as well as some recommendations to improve it. We also provide lot of information about energy uses in our houses as well as a forum to share "energy saving" tips.
I have done summer internships at the following companies:
• Developed customer relations by accommodating all types of
requests (wine sales, event hosting)
• Executed marketing strategy through commercial videos, print and online media
• Analyzed the operational efficiency of the plant by
working eight-hour shifts on the shop floor
• Offered solutions to reduce losses and improve workers’ overall wellness
Concentration: Technology and Engineering Management
Coursework: Machine Learning, Deep Learning and NLP, HCI, Strategy, Marketing, Management and Finance for Entrepreneurs
September 2014 – December 2014
Team members: Louis Eugène, Guillaume Rostaing
We developped algorithms that automatically generate music to
accompany an instrument. For example, given a guitar track, we
were able to generate a drum track that fits well. It helps
musicians create music and improvise in concert with other
How we did it:
• We generated our dataset from midi files of famous songs from a diversity of different registers.
• Based on our knowledge of music, we designed 16 features for our instruments that we later optimized using Machine Learning algorithms.
• Thanks to those features, we designed clusters for each instruments using algorithms such as k-means, mixture of Gaussian, DBSCAN, k-nn.
• Given a new input (say a guitar track), we would assign its closest cluster, look at the features of the other instruments in this cluster (say the drums) and generate a new track based on those features.
January 2015 – March 2015
Team members: Louis Eugène, Alexis Weill, Mansi Kothari, Jocelyn Neff
Grappibulle is a high-end sparkeling grape beverage produced in
France. The goal of this project was to introduce the product in
the US market.
Grappibulle looks like Champaign but doesn't contain alcohol. Just like wine, it is possible to taste the different varieties of grapes. It is therefore a healthier alternative to wine.
What we did:
• We established our go to market strategy: target customers, partners, whole product analysis, and positioning, branding, pricing, and demand creation.
• We completed a customer analysis, competitor analysis, channel analysis as well as a reality test.
January 2015 – April 2015
See "Personal Projects": NotifyMe
January 2015 – April 2015
See "Personal Projects": SmartBox
September 2015 – December 2015
Team members: Louis Eugène, Alexis Weill, Serhan Uslubas, Rahul Venkatraj
See "Personal Projects": SmartBox
September 2015 – December 2015
Team members: Louis Eugène, Guillaume Rostaing, Cyprien de Lichy
GPA: 4.0/4.0 (Top 1% Class)
Concentration: Mathematics and Computer Science
Coursework: Probability, Statistics, Optimization, Algorithm and Programming, Computer Vision, Industrial Engineering
September 2012 – July 2013
See "Personal Projects": SmartDiag
September 2013 – July 2014
See "Personal Projects": OneGlance
Photo collection with exhibits. The photographs represent
objects from everyday life taken in such a way that it is hard
to guess what it is.
• Showcased artistic photographs and developed a website to present and sell them online
• Organized exhibits to showcase the photos in collaboration with public authorities in Cannes and other French cities
Lead studio sessions, taught technical workshops, designed and graded assignments midterms and finals, held office hours.
The skills listed below are a mix of projects I have worked on
(work, startups or development on my free time) as well as classes I
Here is an overview of what I do besides work
You will find my contact information below