Nicolas Gioanni
HomeExperienceResearchProjectsRecommendationsResume
HomeExperienceResearchProjectsRecommendationsResume

Applied AI Research

My research centers on CytoCV and adversarial machine learning, with additional work in computational biology automation.

© 2026 Nicolas Gioanni. All rights reserved except where otherwise stated.

Nicolas Gioanni

Software Engineer

Notices

  • Contact & Communication Terms
  • Site Terms & Accuracy
  • Privacy Notice
  • Security & Responsible Disclosure

Resources

  • GitHub
  • LinkedIn
  • Source Code
  • ngioanni@uw.edu
  • Contact Form

Original portfolio content and media are protected. Site source code is licensed under the MIT License. Third-party names, marks, and materials remain the property of their respective owners.

Applied AI Research

My research centers on CytoCV and adversarial machine learning, with additional work in computational biology automation.

Detail
Showing overview details.
CytoCV Graphical Abstract

CytoCV Demo

5 min 28 sec
Your browser cannot play this video.
1UW Bothell School of STEM logo

CytoCV

CytoCV: Web-based platform for reproducible yeast microscopy image analysis

CytoCV is an open-source platform that turns yeast microscopy stacks into reviewable per-cell fluorescence measurements and structured exports.

Reduced manual analysis by 97%, cut processing time by 73%, and reduced file storage by 56.6%.

  • Translated biology-facing requirements into a repeatable workflow for microscopy analysis.
  • Coordinated the platform work across research, software, and deployment needs.
  • The workflow uses DIC images to identify cells and aligns fluorescence channels for per-cell measurements.
  • Selectable overlays let researchers review analysis results before export.
  • Researchers can upload DeltaVision or TIFF image stacks, monitor analysis, inspect overlays, and download structured data.
  • Protected artifacts and retention controls support repeatable shared-lab use.
  • The public release includes citation metadata and a versioned software DOI.
Live siteSource code
AML Graphical Abstract
2UW Bothell School of STEM logo

Adversarial Machine Learning

An independent study in adversarial machine learning that prototyped four attack families and four defensive strategies across standard image-classification benchmarks.

Optimized batched Keras defense execution to under 5 seconds while preserving 94% clean-data accuracy.

  • The study covered evasion, model extraction, model inversion, and backdoor poisoning.
  • Defenses included adversarial-input detection, input preprocessing, output perturbation, and defensive distillation.
  • Each prototype isolates one threat or mitigation so its behavior can be inspected independently.
  • The benchmark set supports comparison across grayscale and color-image models.
  • Captured experiment metrics and visual outputs to compare the effect of each attack or defense.
  • Kept clean performance visible when assessing whether a mitigation helped.
Source codeReference manuscript
GuideDonorScheduler Graphical Abstract
3UW Bothell School of STEM logo

Guide Donor Scheduler

Guide Donor Scheduler: Yeast DNA target selection automation

Guide Donor Scheduler converts yeast FASTA sequences and requested amino-acid substitutions into CRISPR guide and donor constructs with researcher-ready XLS output.

Reduced manual research workflows by 99%.

  • The tool accepts requested mutations, searches the associated yeast sequences, and prepares candidate designs.
  • Researchers receive an organized spreadsheet instead of assembling each design by hand.
  • The pipeline chooses mutation-adjacent guide candidates and builds the corresponding donor context.
  • Silent changes can be introduced to reduce the chance of Cas9 cutting the edited sequence again.
  • Structured output keeps the source sequence, guide choice, donor design, and validation result together.
Source code