Skip to content
View SandasGH's full-sized avatar

Highlights

  • Pro

Block or report SandasGH

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
SandasGH/README.md

Sanda Puce

I work at the intersection of machine learning, financial risk, and decision systems. I hold an MSc in Artificial Intelligence from King's College London and a First-Class BSc in Computer Science from the University of West London, with a background spanning financial operations and applied data science.

Through building Subtell, I have developed hands-on AI engineering experience across retrieval, bounded asynchronous workflows, and analysis orchestration. I am interested in Data Science, AI Engineering, and Risk Technology roles in financial services.

LinkedIn

Research

My MSc research compared SCARF contrastive pretraining with classical imputation for tabular classifiers under controlled missingness, using financial-risk datasets. The work examines robustness under cell-level and structured missingness as well as limited label availability.

Selected projects

Investigates the robustness of tabular classifiers under controlled missingness and label scarcity. It compares complete-case analysis, median and KNN imputation, and SCARF-style contrastive pretraining, with retained results and explicit reproducibility limits.

Reconstructs a forecasting project around strict temporal evaluation rather than headline performance. The implementation uses lagged features, purged time-series validation, train-only scaling, and tests for key leakage boundaries.

Presents a publication-safe case study derived from selected patterns in a private AI engineering project. The public excerpts demonstrate concurrent multi-store context retrieval and abortable timeouts without exposing product-specific logic or data.

An interactive D3.js project exploring wealth, education, and human-development patterns from 2000 to 2020 through linked geographic, statistical, and temporal views. View the live visualisation.

Education and focus

  • MSc Artificial Intelligence — King's College London
  • First-Class BSc Computer Science — University of West London
  • Focus areas — Data Science, AI Engineering, and Risk Technology

Pinned Loading

  1. scarf-missingness-tabular scarf-missingness-tabular Public

    MSc Artificial Intelligence research comparing SCARF contrastive pretraining with classical imputation under controlled tabular missingness.

    Jupyter Notebook

  2. stock-forecasting-integrity stock-forecasting-integrity Public

    Leakage-aware stock forecasting with purged time-series evaluation, SVR/LSTM pipelines, and a clear separation of historical results.

    Python

  3. subtell-architecture-showcase subtell-architecture-showcase Public

    Publication-safe AI architecture case study covering multi-store retrieval, bounded asynchronous workflows, and tested JavaScript excerpts.

    JavaScript

  4. global-wealth-visualisation global-wealth-visualisation Public

    D3.js visualisation exploring global wealth, education, and human-development patterns through linked map, scatter, and temporal views.

    HTML