Marco K. Carbullido

Research Assistant, Deep Learning, Software Development

• Improved protein sequence embeddings via masked-token prediction for UniRef sequences, and trained on structures via distillation with AlphaFold2. Implemented a novel transformer-based architecture for protein b-factor prediction, achieving state-of-the-art performance (PCC>0.8).
• Successfully developed and implemented a novel bioinformatics pipeline for protein epitope prediction, combining AlphaFold2 predictions with a deep learning model (including transformer architectures) trained on the majority of the Protein Data Bank for b-factor prediction—extending epitope algorithm scope beyond experimental input.
• Architected and developed a web server and accompanying tool implementing state-of-the-art immunology and bioinformatics algorithms for lab and open science use.
• Implemented a scalable distributed RESTful API framework for GPU-accelerated computing of computationally intensive tasks, reducing computation time by up to 180x via cloud services.
• Delivered useful software for labs researching protein structure-based prediction of CD4+ T-cell epitopes and contributed to advancements in deep learning applications for biology.
• Reviewed, synthesized, and communicated the latest scientific literature on protein science, bioinformatics software, and data analysis to cross-functional teams; led group discussions to enhance lab performance.
• Independently managed the design and implementation of data analysis software to optimize internal protein analytics, enhancing lab workflow to support any potential protein sequence.
• Wrote and contributed to scientific publications and technical reports based on original research, presenting findings to multidisciplinary collaborators.
• Collaborated extensively during whiteboard sessions, proposing improvements to CD4+ T-cell epitope prediction algorithms.

Data Engineer, Deep Learning

• Led efforts to analyze large-scale, micrometer-resolution 3D brain scanning electron microscope (SEM) microscopy imaging data using transformer models to study cellular response to energy demand in the ventral nerve cord (VNC). Segmented each neuron and its mitochondria to analyze cellular response to energy demand. Also familiar with patch clamp and other electrophysiology techniques.
• Set up and managed the entire analysis pipeline for instance segmentation of neurons in the Drosophila VNC using scanning electron microscopy (SEM) images.
• Adapted the Flood Filling Network (FFN) approach for 3D vision transformer-based architectures such as SAM2 from Meta to reconstruct sparse neuronal branching and mitochondrial networks for energy density analysis.
• Organized and coordinated a large-scale crowdsourcing initiative, engaging tens of thousands of contributors to generate high-quality consensus datasets for training robust machine learning models.
• Analyzed mitochondrial distribution, network structure, and size across motor neurons to study relationships between neuronal energetics and morphology.
• Segmented and reconstructed neurons in the peripheral nervous system, mapping projections from the VNC for comprehensive structural and functional analysis.

Research Assistant, Computer Vision / Machine Learning Engineer

• Led automation and scaling of data curation for lab experiments, boosting curation speed and throughput by up to 30x and expanding quantity and quality of research data analyzed.
• Developed and deployed custom, combined deep learning models for multimodal (vision + audio) analysis of animal behavior (especially avian mating displays) over thousands of camera trap videos, tracking and annotating complex behavioral datasets.
• Organized feedback and input from different research teams to optimize workflows, keep comprehensive records, and make process improvements.
• Coordinated data pipeline design and system architecture for robust, publication-ready tools to be used by future research cohorts.
• Collaborated closely with academics and researchers, emphasizing interdisciplinary teamwork and technical rigor in all project phases.

The Open Insulin Project, Laboratory Intern

• Provided bench support for strain engineering: media prep, competent cell prep, mini-preps, restriction digests, and selective plating under aseptic technique; kept organized notes for assigned tasks.
• Gained hands-on exposure to ELISA, pipetting, centrifugation, electroporation, western blotting, small bioreactor runs, and incubator usage; assisted with gene-gun setup/cleanup as needed.
• Became familiar with lab automation tools and shadowed setup of basic automated steps within fermentation workflows.
• Assisted with trials comparing DNA uptake methods and helped document the team’s shift toward a simplified electroporation protocol.
• Participated in Slack-based coordination and check-ins around transformation and expression timelines in Pichia.
• Contributed to protocol review discussions for open-source, low-cost insulin production and followed rigorous sterile practices during assigned tasks.
• Maintained a lab notebook for assigned work and observed troubleshooting approaches used by the team.

Cashier

• Maintained a highly sanitary and organized store environment during the COVID-19 pandemic.
• Efficiently handled customer transactions and adhered to all safety and sanitization protocols.

Medical Bioinformatics and Immunology, Summer Program

• Analyzed large public datasets with dedicated bioinformatics tools for gene expression profiling, miRNA targets, protein characterization, and protein-protein interaction networks.
• Learned bioinformatics pipeline optimization, SNP analysis, and best practices for data-driven translational research.

Volunteer

• Implemented a variety of infrastructure improvement and community projects in collaboration with local residents.

Volunteer Tutor/Mentor

• Co-led academic tutoring and mentoring sessions for students to foster academic engagement and interests.