From Abstract Problems to Concrete Solutions

Briefing

Manual processing of iperf logs across global infrastructures introduced inconsistencies and critical delays in SLA validation. I built a high-performance Single Page Application that centralizes all data processing entirely client-side. The solution converts raw logs into actionable intelligence, automating executive and technical report generation in five languages and reducing analysis time from hours to seconds.

Technical Deep Dive

I implemented a sophisticated parsing engine using a State Machine to process multi-stream iperf3 data flows. The application uses advanced Regex to normalize units (SI vs. Binary) and compute link efficiency statistics with an 80% threshold for automatic PASS/FAIL determination. To guarantee data sovereignty, the architecture is 100% stateless; all processing, complex chart rendering with Chart.js, and PDF generation via jsPDF occur in the browser, eliminating infrastructure costs and privacy risks.

Note: Source code is private due to enterprise compliance and non-disclosure agreements. Detailed system architecture and DevOps documentation are available for review.

Briefing

Rogue DHCP servers can cause outages, bad addressing, and traffic diversion within minutes, yet most teams still rely on reactive troubleshooting. I designed Scapy DHCP Sentinel as a proactive local detector: a host-side sensor sends real DHCPDISCOVER probes, captures DHCPOFFER replies, compares responders against a corporate whitelist, and surfaces suspicious servers in an operational UI before they escalate into incidents.

Technical Deep Dive

The architecture cleanly separates the privileged sensor runtime from the web core: Scapy and BPF run on the host for real network-interface access, while FastAPI, SQLite, and the local dashboard stay isolated in a containerized core. I implemented local HTTP ingest, `trusted/suspect/anomaly` classification, auditable persistence for runs and events, a recurring watchdog, and disk spool buffering so temporary core downtime does not erase operational evidence.

Note: Source code is private due to enterprise compliance and non-disclosure agreements. Detailed system architecture and DevOps documentation are available for review.

Briefing

Provisioning a functional EVE-NG lab on Azure manually means repeating the same network, VM, access, and bootstrap steps every time. I structured that flow as infrastructure automation so the environment can be brought up with a minimal topology, static public IP, and initial EVE-NG installation already coupled to provisioning, reducing operational friction for technical labs, proofs of concept, and individual validation environments.

Technical Deep Dive

The solution combines declarative Terraform to create the Resource Group, VNet, Subnet, NSG, NIC, and Ubuntu VM with `custom_data`, plus imperative guest bootstrap through `cloud-init` to update the OS, adjust GRUB, and execute the official EVE-NG installer. I also kept a complementary reprovisioning path via Ansible and a local operational layer with Azure CLI in `lab.sh` for start, stop, and status operations, prioritizing delivery speed without losing control over the lab lifecycle.

Note: Source code is private due to enterprise compliance and non-disclosure agreements. Detailed system architecture and DevOps documentation are available for review.

Briefing

Building the annual timetable for an educational institution is a classic combinatorial optimization problem that often paralyzes coordination teams for weeks. I built GradeOS to turn this chaotic process into a matter of hours. The system resolved a real logistical deadlock, enabling a complete, validated schedule in under 4 hours after weeks of stalled manual attempts.

Technical Deep Dive

I designed a 100% client-side multi-page architecture (MPA), using Vite 5 to keep builds lightweight and modular. At its core is the 'Seed Environment', an assisted rescheduling engine that enforces strict rules and detects conflicts in real time. With no backend required for the demo, I implemented a State Machine logic with LocalStorage persistence, ensuring the app is resilient, fast, and capable of handling 120+ constraint variables simultaneously.

Note: Source code is private due to enterprise compliance and non-disclosure agreements. Detailed system architecture and DevOps documentation are available for review.

Briefing

Sales teams often lose momentum when switching between lead context and structured sales methodologies (SDR/Closer). DNA Clotilde was conceived as a copilot assistant that reduces cognitive load by delivering scripts, objection handling, and real-time meeting audio summaries, while preserving brand voice and tone.

Technical Deep Dive

I implemented an SSE (Server-Sent Events) streaming architecture to deliver a fluid, responsive chat experience with token cancellation for cost optimization. The AI engine uses dynamic prompt engineering to adjust tone, formality, and objectives based on operating mode (SDR or Closer). I also integrated an asynchronous transcription pipeline with AssemblyAI and Vercel Blob, enabling analysis of real conversations and extraction of insights with structured citations (T#) to ensure factual traceability.

Note: Source code is private due to enterprise compliance and non-disclosure agreements. Detailed system architecture and DevOps documentation are available for review.

Briefing

Small manufacturing businesses and ateliers often struggle with fragmented processes and poor financial visibility. I designed and engineered this vertical ERP to centralize the entire operational lifecycle, from order capture and material management to production scheduling and financial close. The system transformed a manual, spreadsheet-driven operation into a structured, scalable digital workflow.

Technical Deep Dive

The architecture adopts a BaaS pattern with Supabase, where I implemented critical business logic directly in the database layer using Postgres Triggers and RPCs (Remote Procedure Calls) to guarantee atomicity in complex order transactions. Security is enforced through an RBAC strategy using RLS (Row Level Security), ensuring full isolation of financial data. For output, I built a server-side PDF rendering engine that generates executive reports and production orders in real time, ensuring document consistency on any device.

Note: Source code is private due to enterprise compliance and non-disclosure agreements. Detailed system architecture and DevOps documentation are available for review.