LIVE PRODUCTION
Security Operations & Automation
SCOPS — Security Operations Automation Platform
Wazuh SIEM Automated Telemetry Ingestion & Incident Triage Gateway
In active SOC environments, analysts frequently suffer from alert fatigue when reviewing hundreds of raw logs daily. SCOPS acts as an automated triage gateway that connects to Wazuh SIEM managers via REST APIs, evaluates risk scores, and routes incidents into High/Medium/Low priority queues with automated NIST SP 800-61 response playbooks.
Architecture Pipeline: Wazuh Telemetry Stream → REST API Ingest → Python Risk Engine → Priority Triage Queue → SOP Incident Playbook
Python
Streamlit
Wazuh SIEM API
NIST SP 800-61
Docker
LIVE PRODUCTION
Applied AI / LLM Benchmarking
PromptMatrix 2.0 — LLM Prompt Stability Suite
Client-Side Prompt Evaluation, Drift Testing & Strict JSON Validation
LLM applications often fail silently when subtle prompt variations cause schema violations or answer drift. PromptMatrix 2.0 systematically evaluates combinatorial prompt matrices across model variants, scoring outputs on schema adherence with zero server-side secret retention (BYOK).
Architecture Pipeline: Prompt Matrix Input → Systematic Variant Generator → Multi-Model Execution (Gemini API) → JSON Schema Validator
TypeScript
Next.js
LangChain
Gemini API
JSON Schema
LIVE PRODUCTION
Edge NLP / Privacy Engineering
SmartExpenseML — In-Browser Expense NLP
Sub-5ms Colloquial Indonesian Currency Normalizer & Text Classifier
Financial transaction notes must remain private. SmartExpenseML executes 100% on the client device, parsing informal Indonesian amounts (e.g., "35rb", "1.5jt", "ribu") and categorizing expenses without transmitting any data to external servers.
Architecture Pipeline: Raw Text Input → Tokenizer Regex → Category Mapping Rule → LocalStorage Sync
Vanilla JavaScript
Regex Tokenizer
HTML5 LocalStorage
LIVE PRODUCTION
SecOps / Data Pipeline
A.R.Y.A. — SOC Analytics & Threat Hunting
Heterogeneous Log Stream Normalization Layer
Consolidates disparate log streams (JSON, XML, CSV) across network endpoints into unified threat hunting dashboards, reducing manual log inspection time by 30%.
Architecture Pipeline: Log Sources (JSON/XML/CSV) → Normalization Layer → Pandas Aggregator → Streamlit Dashboard
Python
Streamlit
Pandas
REST APIs
