À propos de Ilan
Après avoir étudié à Polytechnique puis à Imperial College, j’ai travaillé pendant 3 ans chez Amazon, où j’ai optimisé la logistique en Europe et à l’international. Je souhaite désormais créer une entreprise dans le domaine de la recherche opérationnelle, et je vois les missions freelance comme un bon moyen de mieux comprendre les besoins concrets de mes futurs clients.
Si votre entreprise a besoin d’expertise en optimisation, je peux vous aider à développer rapidement une solution fiable et efficace, adaptée à vos enjeux, qu’ils concernent le transport, l’industrie, la finance ou l’informatique.
J’ai l’habitude de travailler avec des données incomplètes ou ambiguës, et de les rendre exploitables. Côté techniques, mon champ d’expertise est assez large : recherche locale, programmation par contrainte, programmation linéaire (LP/MIP). J’utilise aussi, si nécessaire, des approches d’intelligence artificielle, allant des LLMs au reinforcement learning.
N’hésitez pas à me contacter, même pour un simple échange. Je serai toujours ravi de discuter de votre problématique, et de vous proposer quelques pistes, même si nous ne travaillons pas ensemble dans l’immédiat.
Français
Bilingue ou natif
Anglais
Bilingue ou natif
Accepte de travailler sur site
Paris (jusqu’à 50 km)
Expériences
- AmazonApplied Scientist IIavril 2022 - Aujourd'hui (4 ans et 2 mois)Paris, France• Created a generic in-house fast and versatile Local Search solver in Julia with a novel system of stateless invariants allowing the use of distributed computing to evaluate hundreds of moves in parallel, and reducing the search space by leveraging DeepReinforcement Learning based heuristics.• Adapted lightning-fast shortest-path algorithms from scientific literature to model complex constraints, applied the solver to Amazon's global Middle-Mile transportation network optimization (thousands of warehouses, tens of thousands of trucks,millions of packages every day).• Conceived an LSTM-based Deep Learning architecture to perform a variable-length regression to forecast arrival times and fill rates of trailers in warehouses to assess operational feasibility in what-if scenarios, improving the former model's accuracy by10 pts.• Led to turning an expensive several days, multiple-people process into an AWS-hosted CI/CD pipeline for a fully automated run of a few minutes, that generated dozens of millions of revenue (saving transportation cost and improving delivery speed)compared to the existing solution.• Acquired management experience with full-time employees and interns, officially managing one full-time employee and 2 interns; and leading full-time employees to contribute to my project.• Interacted with a great variety of teams in order to communicate about how the problem is modelled through meetings, papers,presentations and internal scientific conferences. In under a year, that led to the use of my model both in Europe and North America.
- AmazonResearch Scientist Internavril 2021 - octobre 2021 (6 mois)London, UK• Very important ambiguity by defining the scope and leading a high-impact project involving Machine Learning and Robust Optimization applied to Labor Planning, applied to reducing the variablity of rosters for Amazon warehouse employees.• Under the supervision of Dr. Dario Paccagnan (Imperial College London), wrote a paper as a first author reviewed for IJCAI 2022 but ended up being pulled by Amazon for political reasons.• Leveraged labor plans, supply chain topology information and transportation historical data from hundreds of warehouses• Project leader: worked with 2 full-time employees (1 research scientist and 1 data scientist) to bring the project to success andwas the main point of contact for it
- POLYTECHNIQUE MONTREALResearch Scientist Internmars 2020 - septembre 2020 (6 mois)Montreal, QC, Canada• Created a Julia package that is a hybrid solver using Reinforcement Learning to train a Graph Convolutional Network thatworks as a variable-selection or value-selection heuristic in a Constraint Programming framework• As a first author, published a peer-reviewed paper in the 2021 CPAIOR conference (29 citations in under 3 years),proving good results on the TSPTW, the Graph-Coloring and the Knapsack problem. Received the 2020 Polytechnique's BestResearch Internship Award for this work• Source available on an active Github repository (1300+ commits, 160+ stars)
Recommandations
Soyez le premier à recommander Ilan
Contribuez à la réussite de ce freelance en partageant votre expérience de collaboration avec lui.
Ces profils de freelance correspondent également à vos critères
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Formations
- MSc Artifical Intelligence & Machine LearningImperial College London2021Recherche Opérationnelle, Intelligence artificielle
- Cycle ingénieur - X 2017Ecole Polytechnique ´2021Intelligence articifielle, vision par ordinateur