Manuel García

I consider myself a friendly, responsible, companionate and hardworking person. I am very tidy and every day I like to learn new things as well as to improve myself as a person.

I like sports, food and meeting people. One thing I have noticed is that I don't like to waste time, I always like to be doing something, taking advantage of the time to study, work, develop new ideas or improve in my weak points.

I consider myself a full stack machine learning engineer, with experience in the full life cycle of machine learning. I consider myself responsible, compassionate and hardworking. Always trying to take advantage of time, learning and developing new ideas or improving solutions towards clients requirements and excellence.

I have experience in developing and adapting machine learning models, maintaining best practices to maintain the lifecycle of the models. I have experience in the complete flow of the models implementing all the best practices of MLOps, from pipelines to monitoring the models in production to avoid problems like data drift and keep a constant monitoring of the models performance.

I have a lot of experience in handling audios for all machine learning models, applying transformations and processing necessary for STT models. I have experience with Language models.

One of my strengths is model optimization, building services to consume models such as GRPC, putting models into production on services such as Triton server.

Machine learning engineerInitially I started as a full stack engineer, developing apis in back-end, gaining experience in node, react, express, axios. I was also developing python backend, in which I learned to develop in python, use flask and many python modules.

I configured modules for log management with elastic and fluentd.

About machine learning models, I developed adaptations of models for the needs of the company, in which, I worked with Speech to text and Text to speech models, acquiring knowledge to put them into production.

All the modules I developed were dockerized, which gave me enough knowledge to be able to work with Docker and docker-compose being able to dockerize any module.

I also worked with Google Cloud Platform and learned how to use instances, storage service, load balancers, ML engine and many other features offered by this platform.

Everything we work with is in repositories, from there I learned to use Github and Gitlab, so I have enough experience to work with any of these technologies.

I have the facility to adapt models to the needs according to the required problem, making it scalable and easy to implement.

I have experience with react, react JS and react native, I have develop several web pages for internal usage, and android application.

With react JS, I have worked with firebase as data base and storage, this i avoid the use of a server.

actually I am working in a en to end solution, configuring and developing the necesario services for the product under development, Although with the devops role, i do not have much experience, I manage to fulfill the tasks

I am worked in a en to end solution, configuring and developing the necesario services for the product under development, Although with the devops role, i do not have much experience, I manage to fulfill the tasks.

I have experience with docker and docker-compose, all the products developed in my actual work, should be put in a container, include the ML models, for easy deployment.

I have worked, with multiple container's configuration with docker compose and with internal networks for their communication, as is the case with the tensorflow model server and the models, put in the production.

All containers are configured for expose and API-rest or web socket, for realtime inferences.

All the ML models that I have worked on are scalable, because in the product I am working on, I must maintain a good understanding, since they are calls in real time and also have a low response rate. Taking into account that high effectiveness and high efficiency are difficult to achieve.

In servex I have had to design models with structured data, facing challenges for data analysis, building fast and efficient models taking into account the best practices of MLOps.

In vozy I started in loading the STT system, which was adapted, trained and put into production from scratch.

I was supporting the TTS system, finding a better alternative, selecting and training the best model.

I have had to work very closely with the telephone part, which has made me have to learn about it.

I have supported with the creation of modules, data collection campaigns, web applications, audio and data manipulation to achieve a sustainable dataset for the business area.

I have experience with Vertex AI, in deploying models, I have had to deploy several STT models taking advantage of vertex features to speed up inferences.

At Vozy I have been responsible for managing the STT system. I have proposed, designed and built a data collection system to improve the STT service, designing and building a website from scratch for data collection and model improvement. I had to carry out the whole lifecycle of the machine learning model from data collection to production, developing microservices with API rest as well as a GRPC server for the STT system, taking into account response times and model accuracy as two of the most important factors. I had to create pipelines for data management, using tools such as Prefect, MLflow, Kubernetes for scalability.

For many cases of models in which we had to carry out the ML lifecycle, I had to use tools in AWS, such as sagemaker, glue, ECR, ECS in which I took advantage of all the tools to develop, maintain and monitor models using data pipelines from databases in order to train the models.

​

I have also implemented systems to monitor the models in real time in production to avoid data drift and any problems that may arise. One of the biggest strengths I have is to optimize the models to reduce their size and increase their prediction speed and put them on services like triton server to be used in an agile way. I have had to use aws sagemaker to create and implement small internal test models in the company. I also have experience in python development and using docker. Additionally I have had to handle language models which are necessary for the improvement of STT systems.

I have built and applied data augmentation concepts, having to develop from scratch systems that help us improve our data.

Systems and computer engineering