Steven El Kaakati

Tier 2 Customer Service Engineer

Dun & Bradstreet · Full-time

• Provide advanced technical support for 10+ D&B products and Okta Authentication, serving a global user base from small businesses to Fortune 500 companies
• Sustained 100% CSAT across all months and 100% on monthly quality audits; recognized for the fastest average time-to-resolution on the team
• Own identity and access troubleshooting via Okta SSO (SAML/OIDC/OAuth) — validating claims, token lifecycles, MFA, and SCIM provisioning across dev, QA, UAT, and production
• Diagnose REST/API integration issues in Postman, analyzing headers, payloads, authentication, and error codes to restore failing integrations
• Built a Copilot agent integrated with Salesforce and Jira that surfaces case context, identifies missing information, and drafts customer replies — now in use across the support team
• Use Splunk for log analysis during major incident bridge calls, and coordinate global incident response through production outages and after-hours escalations
• Author KCS-aligned documentation — troubleshooting guides, post-incident RCAs, and product FAQs — reducing repeat incidents and accelerating onboarding

Cloud Systems Engineer

DocuXplorer · Full-time

Held concurrently with full-time coursework toward my B.S. in Information Technology at UNF (completed May 2024).

• Deployed and managed 20+ Windows Servers and virtual machines across Azure and AWS, sustaining 99.9% uptime through proactive monitoring, daily backups, and monthly snapshots
• Developed 15+ PowerShell and Bash automation scripts, scaling automation from 0% to 70% and saving the team 20+ hours per week
• Administered 500+ users via Active Directory, Microsoft 365, and Entra ID across distributed teams
• Ensured compliance with HIPAA, SOX, FINRA, and GDPR; contributed to threat protection, system hardening, and Zero Trust network configuration
• Designed Salesforce CRM workflows and KPI dashboards tracking quota performance and service delivery metrics
• Integrated AI/OCR solutions to eliminate manual touchpoints in document processing pipelines

Technical Support Specialist

DocuXplorer · Internship

Held concurrently with full-time coursework at UNF.

• Delivered Tier 1–2 support for Windows environments, resolving software, access, and performance issues for internal and client-facing users
• Managed updates and patches across internal and client-facing servers
• Documented technical procedures and troubleshooting steps, seeding the team's internal knowledge base
• Built hands-on experience with Azure, AWS, GCP, Active Directory, and remote desktop services — the foundation for promotion into the Cloud Systems Engineer role

Technical Project Coordinator — AI & ML

NLP Logix · Internship

• Delivered AI solutions spanning computer vision and NLP chatbots, on scope and on schedule
• Collaborated with cross-functional teams of engineers, data scientists, and stakeholders; facilitated sprint ceremonies, standups, and planning in Jira
• Acted as primary client liaison, translating business requirements into actionable project plans and user stories
• Authored Statements of Work, defined project scope and deliverables, and contributed to risk mitigation strategies
• Supported integration testing, deployment, and user documentation for client-facing AI systems

Ocean Rescue & Beach Safety Professional

County of Volusia · Part-time

• Served as a first responder on one of the busiest beaches in Florida, maintaining constant surveillance and enforcing beach and water-use regulations
• Performed water rescues, administered CPR, and delivered first aid as part of a lifeguard team completing over 3,000 rescues annually
• Provided on-scene emergency care to stabilize injured patrons ahead of advanced EMS arrival
• Operated ATVs and emergency response vehicles to reach incidents across assigned zones

Built a six-VM Windows Server domain environment from scratch to practice enterprise identity and infrastructure administration at production scale. Covers the full stack — domain services, replication, patch management, configuration automation, and container orchestration — across an isolated virtual network with no cloud dependency and no licensing cost.

Hyper-V · Active Directory · Windows Server 2022 · Group Policy · PowerShell · Ansible · Terraform · Kubernetes

• Deployed primary and secondary domain controllers running AD DS, DNS, DHCP, and Group Policy on a lab.local forest
• Configured AD replication, sites and services, and FSMO role transfers between DC01 and DC02
• Stood up WSUS for centralized patch management across all domain-joined machines
• Built a Linux automation node running Ansible and Terraform to manage Windows hosts over WinRM and provision AD objects programmatically
• Ran K3s Kubernetes on a dedicated node with kubectl and Helm for workload deployment practice
• Segmented lab traffic across two virtual switches — an isolated internal network and a NAT network for controlled internet access
• Documented the entire build as an annotated guide explaining the technical function and enterprise rationale behind every command

AssettoServer is a widely used open-source dedicated server with no maintained web UI — every configuration change means hand-editing INI and YAML files over SSH. I'm building the panel that fills that gap, designed to be released back to the community that maintains the server itself.

FastAPI · React · Vite · Python · Docker · REST API · Open Source

• Architected as a single container with no database — server config files remain the source of truth, so the panel never desynchronizes from the underlying deployment
• Built a content scanner that detects installed tracks and cars, handling both single-layout and multi-layout track structures
• Used comment-preserving YAML serialization and atomic temp-and-rename file writes so hand-authored configs survive round-trips through the UI
• Designed a pluggable restart mechanism that defaults to no Docker socket access, avoiding the container-escape risk most panels accept by default
• Scoped into six sequential milestones with a written architecture brief before any code was committed

Job applications routinely make candidates retype everything already present on the resume they just uploaded. ApplyAssist is a Chrome extension that parses a resume locally, builds a structured profile, and maps it onto application forms — with the resume never leaving the machine.

Chrome Extension · Manifest V3 · React · Vite · Claude API · JavaScript

• Parses PDF and DOCX resumes entirely client-side into a normalized profile schema — no server, no upload, no third-party processing
• Layered field detection combining label text, ARIA attributes, and fuzzy heuristic scoring, with confidence thresholds gating auto-fill versus user confirmation
• Bring-your-own-key Claude API integration handles only low-confidence fields and open-ended application questions, with prompts explicitly constrained against fabricating experience
• Manual mapping mode lets users bind any page field to any profile value when detection falls short
• Debugged Workday's Shadow DOM rejecting native HTMLInputElement property setters, and implemented graceful fallback with shadow root traversal

Took a family towing business from no web presence to a fully automated digital operation. I own the entire stack: entity formation, site architecture, DNS, search indexing, and an automation pipeline that keeps the business's local listings active without anyone touching them.

Cloudflare Workers · Claude API · OAuth 2.0 · Google APIs · JSON-LD · SEO · Cron

• Filed Florida LLC formation and built the business's first web presence end to end
• Deployed a mobile-first static site on Cloudflare Workers with GitHub continuous deployment
• Migrated DNS to Cloudflare, configured a custom domain, and verified the property in Google Search Console with a submitted sitemap
• Implemented JSON-LD LocalBusiness structured data aligned to the Google Business Profile for consistent local search signals
• Built a scheduled Cloudflare Worker that generates unique Business Profile posts twice weekly via the Claude API and publishes them through Google's My Business API
• Handled OAuth 2.0 refresh-token authentication server-side with all credentials stored as encrypted Worker secrets
• Extended local presence across Apple Maps and Yelp with consistent NAP data

A data pipeline that turns raw OpenStreetMap geometry into print-ready gallery posters. Three cities — Beirut, St. Louis, Jacksonville — rendered at 300 DPI from real coordinate data rather than traced images, so the geography is accurate down to the residential street.

Python · Pillow · NumPy · SciPy · GeoJSON · Overpass API · Puppeteer

• Queried the Overpass API for road, water, and coastline geometry, exporting GeoJSON per city
• Wrote a Mercator projection layer mapping geographic coordinates into poster pixel space against a configurable lat/lon anchor
• Rendered at 2× supersampling and downsampled to 5400×7200 for true 300 DPI output at 18×24 inches
• Composited road and water layers with NumPy masking, applying a circular crop with a three-tier road weight hierarchy
• Diagnosed tidal rivers being tagged as natural=coastline rather than water polygons, requiring a separate coastline query and flood-fill approach
• Solved ocean rendering — OSM stores no polygon for open sea, so the fill is computed via flood-fill seeded from offshore coordinates
• Composited typography via headless Chrome and exported dimension-locked PDFs for print

Built an AI-powered agent that automates email drafting for customer communications—delivering faster responses with better accuracy and formatting. This project reflects my belief that AI doesn't remove humans from the loop; it frees us to focus on what matters most: solving problems.

AI/LLM · Automation · Email Integration · Knowledge Base

• Removed the manual drafting step from routine customer replies, shifting focus from writing to problem-solving
• Improved response accuracy and consistency through knowledge article integration
• Designed with human oversight—AI augments decision-making, not replaces it
• Currently in active use, continuously refining based on real-world feedback

Designed and operate a segmented home network as a live security lab — not just a deployment, but an environment I actively monitor and tune. Real firewall logs, real intrusion alerts, real false positives to triage.

UniFi · VLANs · Firewall · IDS/IPS · Zero Trust · DNS Filtering · Network Security

• Segmented the network into five VLANs — Core, IoT, Cameras, Lab, and Guest — each mapped to a defined trust zone with explicit inter-VLAN policy
• Configured the ISP gateway in IP Passthrough so the UniFi gateway owns routing, firewalling, and the public address
• Enabled IDS/IPS with enhanced threat signatures and automated signature updates
• Layered DNS-level blocklists and deployed network honeypots for lateral movement detection
• Triaged live IDS alerts and firewall logs, distinguishing genuine threats from benign traffic
• Diagnosed and resolved a geo-blocking rule silently dropping legitimate corporate VPN traffic — a real-world false positive traced through firewall logs to a specific region rule
• Integrated cameras and doorbell into a unified management plane via ONVIF

Built a fully voice-interactive Discord bot that listens in voice channels, detects a wake word, transcribes speech, queries Claude AI, and speaks responses back in real time. Combines multiple AI APIs into a seamless voice pipeline with a live web dashboard for monitoring and control.

Node.js · Claude AI · OpenAI Whisper · Discord.js · Express · Electron

• Full voice I/O pipeline: Discord Opus audio → WAV → Whisper transcription → Claude → TTS playback
• Wake word detection triggers context-aware AI responses directly in voice channels
• Dynamic user profiling — Claude generates per-user context notes to personalize responses over time
• Streams Claude responses sentence-by-sentence to TTS for natural, low-latency audio delivery
• Real-time web dashboard with live transcripts, Q&A history, and voice channel controls via SSE
• Packaged as an Electron desktop app for standalone deployment

Architected and built the entire customer support platform from the ground up. This end-to-end project gave me deep exposure to business process design, stakeholder collaboration, and balancing usability with productivity.

Salesforce · Flows · Lightning · SOQL · Dashboards

• Designed complete case lifecycle—from intake to resolution and reporting
• Built custom Flows to automate routing, escalations, and SLA tracking
• Created email templates for consistent, professional customer communication
• Developed KPI dashboards tracking quota performance, case volume, and SLA compliance across the sales and service teams
• Collaborated with stakeholders to translate business requirements into scalable solutions

Built and maintain a production-grade gaming server for a persistent daily player community. Demonstrates infrastructure management, security best practices, and community building.

Docker · Linux Ubuntu · Cloudflare Tunnel · Bash

• Containerized with Docker for easy deployment, updates, and resource management
• Hosted on Linux Ubuntu server with automated backups and monitoring
• Secured via Cloudflare Tunnel—no exposed ports on firewall
• Tuned server performance to hold a stable tick rate under sustained concurrent load