Principal Engineer2018 - 2019Jampp
Technologies: Amazon Web Services (AWS), Prometheus, Docker, Redis, AWS, Python, Go
- Proposed an alternate architecture and migration plan (from Python to Go) for a Real-Time Bidder that should outperform the current application and should support 0.5M qps with a max latency of 50ms. Included performance analysis and prototyping components in Go.
- Created an architecture for incremental migration of the current app with coexisting Go and Python components.
- Designed and implemented a device ID database for real-time access (1B keys, <5ms, 0.5M qps) to be used inside the Real-Time Bidder.
- Designed and implemented a centralized component for estimating campaigns budget spend in real-time and stop campaigns immediately when the budget is reached.
Consultant for Data Science Program2014 - 2018Fundación Sadosky
Technologies: MongoDB, Hadoop, Python, Java
- Served as the product manager for Palenque, a platform for agriculture-related data portals and software products.
- Developed web scrappers for Palenque: weather station data (15 nationwide and local networks with different web pages/apps) and grain prices. I used Python/Scrappy and MongoDB.
- Organized data science courses for people without a CS background: R for health professionals, NLP for social scientists. Introductory data science talks.
- Consulted on data science projects in the public sector.
Senior developer2012 - 2018Disarmista
Technologies: System Programming, Python, C, C++
- Created a configurable/scriptable engine for detecting security threats in workstations based on the behavior of processes across the system. Used inside a commercial corporate antivirus as a ransomware detector. Included the development of kernel drivers.
- Created an engine for building snapshots of the state of a Windows workstation (filesystem, registry, memory) and detecting security threats by checking hundreds of IOCs (indicator of compromise). Used inside a commercial corporate antivirus.
- Tuned a machine learning algorithm for detecting malware based on static information extracted from executable files.
- Designed and developed a transparent encryption layer for cloud file storage such as Box, Dropbox and Google Drive (files are encrypted for upload and decrypted for download on-the-fly). Developed for Windows and Mac. Included kernel drivers, a credential server and communication components.
- Created video filters for an iOS educational application using ObjectiveC and OpenGL.
Big Data/Machine Learning Consultant2013 - 2014Taringa
Technologies: Machine Learning, MongoDB, Python
- Created a reporting system for an in-house ad server. It needed to handle reports by the hour, day and month faceted by 3 arbitrary attributes.
- Created a module for ranking posts and choosing the top-n by audience and topic based on an ad-hoc designed formula for estimated the growth potential of each post.
- Created a traffic prediction and forecasting module for an ad server using temporal series.
Senior Developer, Head Engineer, Head of I+D2008 - 2012Popego
Technologies: SciPy, NLTK, MongoDB, Java, Python
- Performed research on NLP and classification techniques for personal media consolidation and developed the Popego site backend features. Main tasks: deep text analysis, topic classification, generation of user profiles, and clusterization of users based on interests.
- Created an automated website classifier/analyzer, with online and extensible training of new topics and categories. Developed the backend of the site and the ML features. Included web scrapping for extracting the relevant text on each page.
- Implemented an ad selection algorithm for an in-house ad server, using CTR prediction and user profile matching based on content classification.
- Performed research and developed prototypes for sentiment analysis, NLP and classification of tweets for measuring user reaction on particular topics and products.
- Created a system for generating feature sets and list of possible values for a given list of products and categories. Included a generic web scrapper that searches for pages with products factsheets, extracts and cleans its features/values.
Intern at Microsoft Research2007 - 2007Microsoft
Technologies: MATLAB, C
- Created an online learning algorithm for reducing (nonlinear) harmonic distortion in voice communications with cheap hardware. Produced a prototype, a US Patent, and a research paper.
Senior Developer, Head Engineer1999 - 2004Core Security Technologies
Technologies: System Programming, Smalltalk, C, C++
- Contributed to the design and the development of a corporate workstation security product. Implemented a Windows login replacement, an authentication server and many assorted features of the product, including enforcing access rights and instrumentation of Windows apps.
- Performed research and developed prototypes for applying innovative visualization techniques for the detection of security breaches by visually analyzing large amounts of log files.
- Coordinated a small team for the development of a product for the enforcement of access rights for files, registry keys and network connections on a per-application basis.
- Created a product for enforcing software licenses. It included automatic binary encryption and a fingerprinting schema for uniquely identify each copy of the software. An US Patent was granted for this product.