Work built around real systems.

A restaurant needs a stronger digital presence that makes its menu, identity, location, and customer journey easier to discover and use.

Designed a modern responsive web experience around the restaurant brand, menu presentation, conversion points, location information, and mobile-first interaction.

Next.js application architecture with responsive layouts, structured content, performance-oriented rendering, and conversion-focused page sections.

Developed the landing experience, navigation structure, menu presentation, contact pathways, and reusable visual components.

A clearer customer-facing digital platform suitable for discovery, direct enquiries, and future online service expansion.

Organizations need visibility into application vulnerabilities before weaknesses become exploitable security incidents.

Combined reconnaissance, application testing, vulnerability analysis, traffic inspection, and structured evidence collection.

Black-box assessment workflow covering attack surface discovery, request analysis, vulnerability identification, evidence capture, and remediation mapping.

Security testing using tools and techniques across web application assessment, network discovery, packet analysis, and vulnerability validation.

A repeatable assessment workflow that turns discovered weaknesses into prioritized remediation actions.

Security events become difficult to investigate when network telemetry and detection data are distributed across disconnected systems.

Built an ingestion and monitoring workflow around network security telemetry, detection rules, centralized search, and operational dashboards.

Suricata → eve.json → Filebeat → Elasticsearch → Kibana, with detection rules and investigation workflows layered over the telemetry pipeline.

Configured network monitoring, custom detection rules, log forwarding, Elasticsearch indexing, and Kibana investigation views.

A centralized security monitoring workflow capable of turning network events into searchable investigation data.

Embedding hidden information into model-training pipelines can create subtle artifacts that are difficult to identify through traditional image-level inspection alone.

Investigated whether information embedded during model training could be detected from model-weight representations rather than only from visible host images.

Host-image training → secret embedding variants → trained model collection → model-weight feature extraction → statistical / machine-learning classification.

Prepared comparative model sets, extracted weight-level features, and explored classifier-based discrimination between models trained with and without embedded secrets.

A research workflow for studying whether training-time information embedding leaves detectable signatures in learned model parameters.