Summary
Overview
Work History
Education
Skills
Websites
Certificates And Education
Awards
Timeline
Generic

Samuel Salas

Cambridge,MA

Summary

I'm looking to work at the cutting edge of technology, within an agile and creative team, at the best company in the world. And I'll find a way to do that!

Overview

3
3
years of professional experience

Work History

Sofware Engineer

Tycho AI
08.2023 - Current
  • MIT startup dedicated to the development of fully autonomous drones
  • Coded Obstacle Detection module with model based approaches using OpenCV in C++ and Neural Networks
  • Coded Trajectory Planning module to navigate an autonomous drone through obstacles in real time in complex environments
  • Deployed Neural Networks with TensorRT Engine in C++ from pyTorch models via ONNX exporting, filling missing operations, configuring fastest settings, enabling DLA on a Jetson. Accelerated inference run-time by 5x.
  • Integrated these modules into ROS2 node that gathers data from sensors and runs thread safe functions. Visualized with Foxglove.
  • Improved Octomap's insertion speed of insertions by 4x and quality of point cloud insertions by preventing the library's inherent of data.
  • Developing hyper-realistic simulation environment in Unreal Engine 5 to test modules with ground truth depth in real time using DDS to communicate with ROS2 node.

Machine Learning Engineering Intern

Grupo Ecologico Sierra Gorda I.A.P
02.2022 - 09.2022
  • Non-profit organization dedicated to the conservation of the most diverse biosphere reserve in Mexico
  • Programmed an Encoder in TensorFlow and used it to classify the images into regenerated or eroded
  • This was a creative strategy to overcome a heavily unbalanced data-set and it yielded over 90% accuracy in the best model
  • Evaluated the forest health of Sierra Gorda from 2015 - 2021 using the Encoder and generated maps of forest health, obtaining the trend of the overall state of the natural reserve in the process
  • Used Google Earth Engine's API and KMZ files provided by non-profit organization GESG to build a data-set of satellite images of Biosphere Reserve Sierra Gorda to train Encoder.

Software Engineering Intern

Meta
08.2021 - 11.2021
  • Used NVIDIA's TensorRT plugins and torch.fx tracer to speed up inference run-time of the Transformer Encoder deep learning architecture in Python
  • Obtained around 7x inference speed-up on average, which helped save GPU resources within the company
  • Transitioned internship into a full-time offer.

Education

Bachelor of Science in Computer Science -

National Autonomous University of Mexico
Mexico
01.2022

Skills

  • Python, C / C, Bash
  • Git
  • Linux
  • Docker
  • DevOps (CD/CI, linting, testing, etc)
  • Computer Vision
  • Machine Learning / AI / Deep Learning
  • Neural network run-time acceleration (TensorRT, ONNX)
  • Obstacle Detection
  • Trajectory planning
  • Simulation
  • Unreal Engine 5
  • Autonomous Vehicles
  • ROS2
  • Parallel computing (multi-threading /thread safe run-time)
  • Foxglove visualization

Certificates And Education

  • Building Cloud Computing Solutions at Scale, Duke University, 2023
  • Deep Learning, DeepLearning.AI, 2020
  • Fundamentals of Software Architecture for Big Data, University of Colorado Boulder, 2023
  • Computational Neuroscience, University of Washington, 2023
  • The Pragmatic Programmer: A. Hunt & D. Thomas
  • Fundamentals of Computer Vision: A. Torralba, P. Isola & W. Freeman

Awards

  • ICPC Regional Finalist, 01/2022, 07/2020, 11/2019, Qualified for the regional finals three times for the region of Mexico (participated three times).

Timeline

Sofware Engineer

Tycho AI
08.2023 - Current

Machine Learning Engineering Intern

Grupo Ecologico Sierra Gorda I.A.P
02.2022 - 09.2022

Software Engineering Intern

Meta
08.2021 - 11.2021

Bachelor of Science in Computer Science -

National Autonomous University of Mexico

Similar Profiles

CREATE PROFILE
Samuel Salas