Experience

IRB Barcelona

• Spain
• Research Services
• 300 - 400 Employee

• Bioinformatician

  • Jun 2023 - Present

  Studying new ways of producing bioactivity singnatures for uncharacterized compounds by fine-tuning molecular representation models to produce bioactivity signatures, and using embeddings from molecular representation models as inputs to neural networks trained on bioactivity signatures from 1 million compounds with experimental data. The primary goal is to make better models for characterization of compound databases in the 25 bioactivity space of the Chemical Checker (https://chemicalchecker.org/) for use in drug discovery tasks. A second goal is for these models to be able to distinguish between stereoisomers of the same molecule, as existing models trained on morgan fingerprints are "blind" to differences between stereoisomers. Supervised a maths and physics student during their summer internship. Conducted high-performance computing tasks on GPU and CPU clusters using Slurm. Show less



IBMB

• Spain
• Research Services
• 1 - 100 Employee

• Bioinformatics internship

  • Jan 2023 - Jun 2023

  Interpreting protein language models (transformer models trained on protein sequences) to understand how protein language models work as well as they do, and see what they can teach us about the "language" of amino acids, with the goal of improving our ability to do de-novo design of proteins. I focused on attention analysis, where we study the weights from the self attention mechanism in transformer models. My main contribution was an interactive tool for visualizing attention patterns on protein structures from 4 important protein language models (ProtT5, ZymCTRL, ProtBert, TapeBert). The tool is built with Streamlit, does interactive visualization of protein structures with STMol and Py3DMol, and is deployed as a Huggingface space with a community grant for a Nvidia T4 GPU. The tool is available at https://huggingface.co/spaces/AI4PD/hexviz/ try it out! The idea behind the tool is that by visualizing the attention patterns of protein language models on protein structures we’ll be able to better use our structural understanding of proteins when interpreting attention patterns, so that we can investigate what biological features protein language models focus on and "understand" when making their predictions. The work was done at the AI for protein design group at the Molecular Biology Institute of Barcelona. Read more about the group at https://www.aiproteindesign.com/ Show less



Oda

• Norway
• Consumer Services
• 400 - 500 Employee

• Software Engineer

  • Jan 2021 - Sep 2022

  Software engineer doing full stack work web development work in the procurement team. Primary language and framework used is Python and Django, deployed using Docker and Kubernetes. Main tasks are: • Designing and implementing systems to do data integration with suppliers using Pydantic and Argo Workflows. • Building and maintaining purchase order flow web tools to help visualize the state of our purchasing systems. With the goal of making ergonomic user interfaces that allow purchasers to be self served in debugging and resolving errors during sending, receiving and processing of purchase orders and supplier deliveries. Built using Django with a Bootstrap UI. • Debug issues with our own systems and third party integrations. Using Datadog, sentry and production shells. Setting up alerts and monitors to ensure errors are detected and corrected early. • Maintaining the data intensive integrations with Oda's purchase orders forecasting system • Extended the internal python service to service communication library built on top of RabbitMQ to be more ergonomic and flexible when used with Django, simplifying the definition of new queues and messages between services. • Designed and set up events and consumers for moving a 1-1 service communication to a 1-n communication, using the internal service to service library. • Maintained and developed new features for the inventory systems in Oda's warehouses using React, GraphQL, Django. Show less



Blueye Robotics

• Norway
• Maritime Transportation
• 1 - 100 Employee

• Software engineering intern

  • Jun 2018 - Dec 2020

  Developing a open source Python SDK for interfacing with a underwater drone. Formalizing the TCP communication between underwater drone and app. CI tests integrated with hardware using Pytest. Designing and developing a battery cycling test setup. C++ driver development and testing for the Waterlinked A50 DVL (Doppler velocity log). Developing a open source Python SDK for interfacing with a underwater drone. Formalizing the TCP communication between underwater drone and app. CI tests integrated with hardware using Pytest. Designing and developing a battery cycling test setup. C++ driver development and testing for the Waterlinked A50 DVL (Doppler velocity log).



Oda

• Norway
• Consumer Services
• 400 - 500 Employee

• Software development intern

  • Jun 2020 - Aug 2020

  Developing a web app for performing inventory correction at a warehouse with a focus on user testing and a user interface that would help reduce human error when performing inventory corrections. Technologies used: Django backend , React + typescript frontend with graphQL. Got to extend my internship for 3 weeks to work integrated in the inventory team at Kolonial.no working alongside them on their issue board, main task was developing a dashboard for giving a daily overview of operations at a warehouses incoming terminal. Show less