Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I'm Christian Cancedda, a researcher and engineer blending AI, machine learning, and software development to create deployable products. I enjoy turning novel ideas into scalable solutions—from explainable recommendations to robust visual localization pipelines. I collaborate across academia and industry, translating complex research into practical impact, and I thrive in ambiguous environments where teamwork and curiosity drive results.…I'm Christian Cancedda, a researcher and engineer blending AI, machine learning, and software development to create deployable products. I enjoy turning novel ideas into scalable solutions—from explainable recommendations to robust visual localization pipelines. I collaborate across academia and industry, translating complex research into practical impact, and I thrive in ambiguous environments where teamwork and curiosity drive results.

Christian Cancedda

Data Scientist, AI Engineer, Developer, +2





I'm Christian Cancedda, a researcher and engineer blending AI, machine learning, and software development to create deployable products. I enjoy turning novel ideas into scalable solutions—from explainable recommendations to robust visual localization pipelines. I collaborate across academia and industry, translating complex research into practical impact, and I thrive in ambiguous environments where teamwork and curiosity drive results.…I'm Christian Cancedda, a researcher and engineer blending AI, machine learning, and software development to create deployable products. I enjoy turning novel ideas into scalable solutions—from explainable recommendations to robust visual localization pipelines. I collaborate across academia and industry, translating complex research into practical impact, and I thrive in ambiguous environments where teamwork and curiosity drive results.

Available to hire

I’m Christian Cancedda, a researcher and engineer blending AI, machine learning, and software development to create deployable products. I enjoy turning novel ideas into scalable solutions—from explainable recommendations to robust visual localization pipelines.

I collaborate across academia and industry, translating complex research into practical impact, and I thrive in ambiguous environments where teamwork and curiosity drive results.

Skills

Python

PyTorch

C++

Experience Level

Python

Expert

PyTorch

Expert

C++

Expert

Language

English

Fluent

Italian

Fluent

Work Experience

Machine Learning Engineer at Blue Reply

April 1, 2019 - September 30, 2019

Research and development of a recommendation system for insurance policy recommendations based on users’ socio-economic data.

Researcher - Covid@Lombardy at CERN / CS4OD collaboration

November 1, 2020 - February 28, 2022

Collaborative research effort on understanding the socio-economic factors that determined the difference in spread of Covid-19 across European regions.

Research fellow - LLM-based recommender systems at University of Cagliari - Mirko Marras Lab

September 1, 2022 - June 30, 2023

Two papers published on Explainable recommendation research to increase user trust metrics.

Research fellow - Visual-geo localization at Polytechnic of Turin - VANDAL group

June 1, 2023 - November 30, 2023

Built, validated and deployed a large-scale search-engine for Visual-localization at street level granularity: 70% of test images localized within 1km radius and 85% within 10km. Calibrated to achieve millisecond-query times. Further details are under NDA.

PhD student - Visual-geo localization and 3D reconstruction at Polytechnic of Turin - VANDAL group

November 1, 2023 - November 30, 2024

Worked on AI-based visual geo-localization and on image retrieval component for 3-D reconstruction (COLMAP pipeline).

Tech Lead & Research engineer - Transferable physical adversarial attacks at Capable

November 1, 2024 - Present

Research to product of an AI camouflage t-shirt for privacy, that is effective against YOLO-like object detector ensembles.

Web Captchas that break AI agents at Stealth startup

August 1, 2025 - Present

Created an AI model that generates ReCAPTCHA-like challenges. LLM Agents’ solve rate on the challenge drops by 70% compared to Google’s ReCAPTCHA. Tested for joint effectiveness on LLMs from OpenAI, Anthropic, Google, Gemini, Mistral, Grok.