Caio Krauthamer Email & Phone Number

Tech Lead and Software Engineer on Machine Learning Projects.Python Developer.FastAPI Framework.Pandas and Numpy Packages.Postgre Database.On Prem Deployment.Monitoring / Observability with Datadog.Main Accomplishments:- Restructure multiple projects to make them production ready.- Rethink the whole CI and CD pipelines to allow easy deployments and enforce code quality.- Centralize common tooling into a private python package shared by multiple projects.- Guide the team for better logging standards and create Datadog dashboards to allow observability over the services allowing the team to quickly identify failures.- Oversee projects from 0 to 1.- Build TRDs and BRDs documents for major refactors and new projects.- Migrate projects and services from bare metal to dockerized runs, allowing better resource management.

São Paulo, São Paulo, Brasil

Software architect and engineer of investments consolidation platform.Python and Typescript Developer.Django and React Frameworks.Postgre Database.Deployment using AWS.Monitoring / Observability with New Relic and Cloud Watch.Main Accomplishments: - Worked on new custodian integration (reverse engineering and web scraping). - Updated code architecture into Micro Kernel Architecture to easily accept new components.

Chicago, Illinois, Estados Unidos

Architect and developer of micro-services.Python and Javascript/Typescript developer.Flask, SQLAlchemy, Alembic, Apollo (GraphQL) and React (components logic only, no styling/CSS) frameworks.MySQL (MariaDB and Postgres) database.Main Accomplishments: - Reworked whole back-end APIs to make code more standardized and reduce duplication, hence making it easier to maintain. - Added strict API validation layer, with descriptive error messages, which reduced amount of time other teams had to spend with us to understand what was wrong with their inputs. - Enhanced a huge amount of APIs response times, in some cases reaching 10x speed ups - Documented all APIs into OpenAPI, helping QA team to automate their test and ease integration of other micro-services with ours. - Exposed all APIs into GraphQL. - Update Front-End code whenever needed to consume new APIs (given that pages where already styled).

Campinas, São Paulo, Brasil

Architect and developer of micro-services.Python and Javascript/Typescript developer.Twisted, Django, FastAPI, SQLAlchemy, Alembic, Apollo (GraphQL), React + Redux (components logic only, no styling/CSS) and Prometheus frameworks.MySQL and MongoDB databases.Deployment using Docker.Main accomplishments: - Facebook Messenger API integration to an omnichannel micro-service utilizing Facebook API. - Implementing a new GraphQL micro-service using Apollo library to handle all communication between the front-end and back-end. - I lead a front-end revamp, in means of design and logical architecture, in which we made a poor use of React + Redux when it comes to performance. After revamping it, the users were able to have a faster and more fluid experience, measured by researches we done with our main clients. - I lead a back-end revamp from a OO (object oriented) architecture to a Event Sourcing using DDD (Domain Driven Design) + CQRS (Command Query Responsibility Segregation). This micro-service had huge response times, and by the end of the revamp, the response times speeded up by 5x under heavy load, and it also improved concurrency, allowing us to scale our product to more simultaneous users, using the same server hardware.

Campinas E Região, Brasil

ACTSR Project - Algorithms Acceleration for Text Spotting and Recognition:The research done covered two libraries for spotting text and one for recognize it in C++.The team was divided into two sub teams: the one that I participated in was focused in optimizing the libraries for text spotting (libccv and OpenCV).The images test set to compute the implementations speed up and accuracy was ICDAR 2013, remarkable competition on text spotting. Across the project, to test our code, I used a specific hardware, Artik 530, and it was required not only to learn about cross compile techniques and GBS emulator but also how to manage all the network to send codes and inputs and receive its output.For the spotting sub team, our main goal was to optimize execution time, and secondarily to boost the accuracy. We had different approaches in each library: - OpenCV: Our main result was to use gray scale images, rather the the default RGB; Parallelizing through OpenMP; Modifying the algorithm to detect full sentences (but keeping the functionality to detect each word separately, if requested); And applying a post processing retry policy to enhance accuracy, without harming too much the execution time. We managed to speed up the code by 6.448x and increasing the F-Score (accuracy) utilizing IoU of 0.5 from 0.162 to 0.480. - libccv: We only parallelized a filter in gray scale and its negative channels. Also we came with a new data structure that could be shared by text spotting and recognition, reducing and overhead to migrate from a data structure from spotting to recognition. We managed to speed up the code by 6.363x, keeping the same accuracy.

Campinas E Região, Brasil

Computer Algorithms in C Monitor.I solved weekly small projects that the students would take later on during the semester so the teacher could fine tune its details and I/O tests.Also, I helped students with their doubts regarding those projects and general algorithm exercises.