Skip to content
View subrosa5's full-sized avatar
〰️
〰️

Block or report subrosa5

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. 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
subrosa5/README.md
subrosa — Machine learning and applied cryptography

I build ML applications and explore applied cryptography, with a focus on evaluation, failure modes, and reproducible experiments.

Working with Python · scikit-learn · XGBoost · PostgreSQL / pgvector · Docker · Go

Selected work

01 / MACHINE LEARNING

Hybrid retrieval, reranking, source checks, and an MCP interface. An evaluation suite covers answers and refusals.

Python pgvector RAG MCP

02 / APPLIED CRYPTOGRAPHY

Streaming encryption, shared key recovery, and a synthetic crash lab. Experimental; no independent security audit.

AES-GCM Shamir ML-KEM ML-DSA

03 / MODEL EVALUATION

Leakage-aware fraud modeling, cost-based evaluation, SHAP analysis, and adversarial robustness experiments.

scikit-learn XGBoost SHAP

04 / SYSTEMS

A Go implementation of Raft with a write-ahead log, crash recovery, and a cluster dashboard.

Go Raft WAL

Exploring

Privacy-preserving ML · encrypted inference · post-quantum protocols

More projects

Εν οίδα ότι ουδέν οίδα

Pinned Loading

  1. antifraud-detector antifraud-detector Public

    Fraud detection ML pipeline: leakage-safe modeling, cost-based evaluation, SHAP interpretability, code review, and black-box adversarial/robustness testing.

    Python 1

  2. wazuh-detection-lab wazuh-detection-lab Public

    Python 1

  3. 0xc12d5d4abbb8c1bee4a5e4ce 0xc12d5d4abbb8c1bee4a5e4ce Public

    Experimental offline threshold file vault: AES-256-GCM, protected Shamir shares, crash lab and reproducible reports.

    Python

  4. local-coding-copilot local-coding-copilot Public

    Pre-enterprise POC for a closed-loop local coding assistant. RAG/MCP demo: subrosa5/rag-mcp-agent. Synthetic Python fixture: subrosa5/exchange-demo. No cloud AI; model/GPU validation pending.

    Python