Pablo Woolvett Email & Phone Number

Concón, Valparaíso, Chile

I worked in projects which deliver computer vision-based information toindustry processes to aid in ops and security.On each of the projects, i've worked in several areas: Hardware Design, SoftwareIntegration (python, Gstreamer), Transfer leaning (CVAT, nvidia TLT),backend (fastapi, kafka, postgres), and frontend/UI (Kivy, Vuejs).One of the projects consists of delivering environment-aware information using AI toforklifts used in a mining facility. Video and depth streams are collected using IntelRealsense cameras, and processed in realtime using a Gstreamer pipeline. The pipelineuses Nvidia's Deepstram plugins for the AI models (Object Detection / segmentation).The software is run from Jetson Xaviers, mounted on each forklift.The whole codebase uses python (and cpp-compiled extensions where possible). Theresulting information is delivered to the user through touchscreen-based UI and forklift hardware (speakers and brakes, in case of emergency).Another project consists of delivering security information (eg ROI occupation, PersonalProximity) using Computer Vision models from several cameras to safety personnel. Livefeed form (RTSP) cameras is fed to a GStreamer pipeline which uses Deepstream to extractper-frame detection metadata related to potential hazards. The pipeline runs in anon-prem centralized server, which then further analyzes and showcases video streams inthree views: live cameras, per-event (as produced by analysis), and security dashboard.

Santiago Province, Chile

I worked building NLP models for Purchase Orders classification, helped structuring catalogue data, and designed/implemented micro-services for text-based predictions and real-time auto-completion.The roles ranged from defining client-based requirements, API design/implementation, frontend design/implementation, CI, CD, to devops. All work were performed following a loose, agile-inspired workflow.The core of my work consisted of deploying specific micro-services to make CRUD queries from ORM models (postgresql, sqlalchemy), deliver autocompletion functionality (elasticsearch-dsl), and NLP-based, supervised and clustering predictions (ludwig, keras). The APIs (fastapi, flask, django) used oauth-based authentication (AWS cognito) and contain self-documented code: both internal (sphinx, napoleon) and external (openapi, swagger). In order to consume the API (and fulfill the User Stories), I also developed framework-agnostic frontend elements (native js web components).In order to build the NLP models, several web catalogs/portals were constantly crawled (scrapy, beautifulsoup), in order to enlarge and diversify datasets. Stuctured ETLs (jinja, sqlalchemy) were performed afterwards to ensure dataset integrity and consistency. Initial inspection was performed using standard data science tools (jupyter, altair) and methods (T-SNE, PCA, SVM, CNNs). Development considered unit/integration (pytest, hypothesis) and behavioral (behave) testing, and was connected from git with Continuous Integration and Deployment (azure devops).CI: Programmed to enforce standard code in both format (black, docformatter) and linting (flake8, pylint, mypy), report code coverage (coverage.py), and follow 12-factor based principles (poetry, pydantic, dotenv, docker). A successful CI build ended with a dockerized, scalable, documented, versioned, 12-factor-friendly API.CD: APIs were deployed to dev/qa/prod, following both automated and manual gates, and considered automatic rollbacks.

Ñuñoa, Santiago

I worked in a variety of projects: scraping, RPA automation and IOT.The scraping project consisted of gathering data from several crawled sources (scrapy) and consolidating it with publicly available databases.The RPA project consisted of synchronizing bank account statuses and transactions (requests+beautifulsoup) with a thirdparty's ERP (using uipath).The IOT project consisted of controlling a photovoltaic battery inverter connected to pV pannels, a battery, and the electric grid. The device was controlled using a on-site connected raspberry pi, which forwarded (via zmq) commands and status data to/from an Ionic webapp.

Viña Del Mar, Chile

Founding member of pikd, an Austrian-Chilean startup.I was part of the team in charge of building the main product, and worked throughout the development of the whole project, mainly in the following areas: data acquisition, labeling and database generation, Machine Learning, and Android development. This project had as a goal to facilitate the generation of high quality content from point-of-view footage from GoPro (and similar) cameras.In order to do this, I worked applying Machine Learning techniques to inertial sensor data associated with the footage in order to detect, classify, and rate associated sport events.I also worked developing two Android apps which show the functionality and capabilities of such developed models, designing the UI according to Material Design guidelines, and implementing the Mobile Vision API (both real time and on-demand), for image processing.Finally, I studied the migration of the already developed models to Deep Learning architectures in order to improve performance and facilitate automation.

Valparaíso Province, Chile

Given a electronics board designed for RFID (Radio-frequency identification), I worked in the re-design of the antenna and its adaptation circuitry in order to improve the reach of contact-less cards. The boards were already placed in position, so the re-design was performed according to specific space and power constraints.

Valparaíso Province, Chile

I worked developing scripts for Matlab which obtain the optimal linear beam-forming matrices in wireless MIMO (multiple input, multiple output) transmissions with per-antenna power constraints.

Magister En Ciencias De La Ingeniería Electrónica, Electrical, Electronics And Communications Engineering

Ingeniería Civil Electrónica, Electrical, Electronics And Communications Engineering

Details of this degree can be seen at: https://usm.cl/admision/carreras/ingenieria-civil-electronica/