Davide Z.

Key Skills

Software

Other

Don't disclose

Top Subject Matter

Healthcare/Drug discovery (Type 2 Diabetes)

molecular property modeling

AI for Science / Drug design

Top Data Types

Text

Document

Medical Dicom

Top Task Types

Text Generation

RLHF

Freelancer Overview

Brings 5+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other and Don't disclose. Education includes Master of Science, Umeå University (2023) and Bachelor of Science, University of Science, VNU-HCM (2021).

Labeling Experience

Controllable Agentic AI: a survey — ongoing project

OtherText

Conducted a PRISMA-style conceptual survey on controllability in agentic AI systems. Reviewed guardrails and constraints approaches and how they interact with adaptive reinforcement learning and agent-in-the-loop workflows. Synthesized human-in-the-loop and agentic controllability mechanisms into an organized survey narrative. • Performed PRISMA-style systematic conceptual review. • Summarized guardrails, constraints, and adaptive RL mechanisms. • Covered agent-in-the-loop and human-in-the-loop approaches. • Produced structured survey findings suitable for publication.

2025 - Present

Multi-objective Optimization in Drug Design — ongoing project

Don't discloseTextRLHF

Developed Transformer-based generative modeling and Pareto-guided reinforcement learning methods for multi-objective ADMET optimization in de novo drug design. Built a standardized benchmark/leaderboard to evaluate multi-objective generative performance under fixed computational budget constraints. Investigated hybrid weighted-sum/Pareto formulations and objective subset selection strategies for scalable high-dimensional optimization. • Designed model training approaches for multi-objective ADMET optimization. • Created and maintained evaluation benchmarks/leaderboards for drug design. • Studied curse-of-dimensionality impacts under budget limits. • Explored objective subset selection and hybrid optimization strategies.

2025 - Present

Predicting Effective Treatments for Type 2 Diabetes — course project

OtherTextText Generation

Worked on QSAR-based analysis and modeling for diabetes treatment-related molecular properties using data-driven feature extraction. Applied PCA and PLS methods to interpret molecular datasets and support predictive treatment relevance. This project involved preparing and transforming structured molecular/property data into modeling-ready representations. • Applied QSAR (Quantitative Structure-Activity Relationships) feature analysis. • Used PCA for dimensionality reduction and interpretation. • Used PLS (Projection to Latent Structures) for predictive modeling. • Evaluated model performance to achieve a high course outcome.

2022 - 2022

Education

Work History