Computer Science student at Arizona State, focused on cybersecurity. I build networks and security tools in my home lab and certify what I learn. The projects and credentials are below.
I'm Sam, a Computer Science student at Arizona State studying cybersecurity. I graduate in December 2026 with a 3.90 GPA.
What I care about most is cybersecurity, IT, and AI, and I try to keep up with how each of them is changing. I like learning new things and getting certified as I go. I've earned nine certifications across security, networking, cloud, and systems, and the full list is below.
Right now I work in tech support at ASU, where I help faculty, staff, and students with hardware, software, and network problems, image machines, and handle Active Directory and ServiceNow tickets. Before that I spent a summer as a software engineering intern at Globule Technology, building the front and back end of a web app with Angular and Spring Boot.
A lot of my learning happens outside of class. I run a home lab on a Proxmox cluster with a pfSense-segmented network, where I practice routing, threat hunting in Splunk, and incident response. On my own I've also built an ML-based SIEM that flags attacks in log data, an LSTM model that catches anomalies in network traffic, an agentic AI assistant that runs its own tools, and a C++ arcade game. They're all in the projects below.
On the technical side I'm comfortable with Python, C/C++, and Bash, a good chunk of networking and security tooling like Wireshark, Splunk, Metasploit, and Burp Suite, and cloud and systems work across AWS, Linux, and Windows. The full list is in the skills panel.
I'm open to roles in cybersecurity, AI, and IT.
A self-hosted lab on a Proxmox VE cluster that I use as a full-scale practice ground — a segmented network, the complete routing-and-switching stack, and a live security-operations pipeline running on top. It's where I turn the things I study and certify into something I've actually built, broken, and defended.
Stood up a multi-node lab on a Proxmox VE cluster and segmented it behind a pfSense firewall — VLANs, DHCP scopes, and firewall policies enforcing hard traffic boundaries between zones.
Provisioned Kali and Ubuntu VMs alongside Docker containers to serve as attacker, target, and service hosts across the environment.
Configured and tested every major routing protocol across virtual routers — OSPF and IS-IS (link-state), RIP and EIGRP, and BGP (path-vector) — verifying reachability and path selection across the topology.
Spent as much time breaking it as building it: diagnosing why neighbor adjacencies won't form or routes won't propagate, then tracing each failure back to the misconfiguration and fixing it.
Layered enterprise switching on top: VLAN trunking, inter-VLAN routing, spanning tree (STP), EtherChannel aggregation, ACLs, and QoS policies.
Deployed Splunk as the SIEM and ran threat hunting and incident response against the BOTSv2 dataset, pivoting through packet captures in Wireshark to reconstruct attacker activity.
Built custom detection alerts and dashboards to surface, correlate, and visualize security events across the lab.
A Python SIEM pipeline that ingests security logs, engineers eight features per time window (IP entropy, error ratios, failed logins), and uses an Isolation Forest to catch brute-force, port-scanning, DDoS, and privilege-escalation attacks — with SOAR-style automated response playbooks, severity-tiered alerting, and a four-panel analytics dashboard.
↗ 02An agentic assistant (Gradio + a custom MCP server) where an orchestrator picks its own tools via OpenAI function calling. It answers questions about my background using RAG over my PDF/resume (embeddings + retrieval), routes security evidence to four defensive agents — log triage, phishing, config audit, and a CVE agent that runs live web search — and takes real actions: simulated Nmap/ping, browser navigation (Playwright), and Pushover contact alerts. End-to-end tool use, not just chat.
↗ 03A TensorFlow/Keras LSTM trained on normal traffic (10-step lookback, MinMax scaling, Adam, 80/20 split) that forecasts what should come next and flags error spikes above a mean+3σ threshold as anomalies — catching injected DDoS and exfiltration, scored with precision, recall, and F1.
↗ 04A supervised ML pipeline (XGBoost) that flags illicit transactions in the Elliptic Bitcoin dataset — 203K real blockchain transactions labeled licit/illicit. The focus is the hard part of fraud detection: severe class imbalance, handled with cost-sensitive learning and SMOTE and benchmarked against a no-handling baseline. A temporal train/test split mirrors real deployment, and results are scored on PR-AUC and precision@top-k instead of misleading accuracy.
↗ 05A dig-and-survive arcade game built on a hand-rolled object-oriented actor system that drives every interaction between the player, NPCs, and a destructible environment — player controls, digging, resource collection, and combat. Enemies navigate with BFS pathfinding, line-of-sight detection, and event-driven behavior.
↗ 06A Naive Bayes text classifier that flags spam SMS messages, reaching 91.3% accuracy on the SMS Spam Collection dataset (5,574 messages). I handled the NLP preprocessing in NLTK (tokenization, stemming, stop-word removal) with CountVectorizer for features, and tuned the model with GridSearchCV using 5-fold cross-validation and F1 scoring.
↗Certificates are issued under my legal name, Sinan Adalioglu — I go by Sam.