Cybersecurity and machine-learning researcher building evidence-grounded systems for trustworthy security decisions.
SecOpsAI · Documentation · Research blog · ORCID · Publication · LinkedIn
I am a cybersecurity researcher and engineer based in Istanbul, Türkiye. I hold an MSc in Cybersecurity from Üsküdar University, completed with High Honour, and a BSc in Biochemistry from the University of Port Harcourt.
My work combines malware analysis, software-supply-chain security, machine learning, security operations, and the design of systems that keep evidence, uncertainty, and human responsibility visible. I am the sole creator of SecOpsAI, a local-first platform for security telemetry, investigations, package research, detection learning, and controlled publication.
- Evidence-grounded generative and agentic AI for cybersecurity
- Malware classification and software-supply-chain threat analysis
- Causal, explainable, and uncertainty-aware machine learning
- Knowledge representation, provenance, and formal reasoning for reliable systems
- Human oversight, abstention, auditability, and rollback in automated decisions
- Privacy-conscious and local-first security architecture
These are active research directions. I distinguish them from capabilities already validated through completed experiments or publications.
| Component | Purpose |
|---|---|
| SecOpsAI Core | Evidence-first triage, package research, asset graphs, detection workflows, agent telemetry, and guarded automation. |
| Mission Control | Operator interface for findings, investigations, research cases, automation, disclosure, and reviewed publication. |
| SecOpsAI Edge | Local sensor for authorised asset discovery, risky-service detection, change tracking, and normalised Core ingestion. |
| Research workflow | Safe artefact intake, static analysis, package comparison, evidence provenance, IOC review, disclosure controls, and publication gates. |
SecOpsAI treats model output as analytical assistance, not evidence. High-impact actions remain bounded by explicit approval, audit records, and reproducible inputs.
| Project | Focus |
|---|---|
| Malware Detection on an Optimised Edge Device | Transfer learning, model optimisation, ONNX inference, and low-latency Android deployment. |
| Deep Learning Malware Classification | CNN-based malware-family classification and experimental evaluation. |
| AI-Powered SOC Agent | Bounded natural-language security analysis and structured incident reporting. |
| AWS Bedrock RAG Project | Terraform-deployed retrieval workflow using Bedrock, Aurora PostgreSQL with pgvector, S3, and scoped IAM. |
| SecOps Autoresearch | Controlled detection optimisation with evaluation, feedback, and rollback-oriented workflows. |
Agada, O. C. (2025). “From Training to Edge Inference: A Transfer Learning Pipeline for Low-Latency Malware Classification on Snapdragon-Enabled Samsung Galaxy S23 Ultra (Android 13).” Journal of Scientific Reports, 11(1), 51–64. DOI: 10.58970/JSR.1135
My MSc thesis, “A Comparative Analysis of Transfer Learning Approaches for Malware Classification Using Convolutional Neural Networks,” compared transfer-learning architectures on malware-family datasets using accuracy, precision, recall, F1-score, AUC, confusion matrices, class-imbalance analysis, and generalisation checks. YÖK National Thesis Center Reference No. 10700043.
- Preserve the source and provenance of every material claim.
- Separate observed facts, reasoned inferences, model suggestions, and unsupported conclusions.
- Never execute untrusted packages on an operator or application host.
- Require explicit approval for sandbox execution, disclosure, mitigation, and publication.
- Evaluate detection changes on holdout data before controlled activation.
- Record negative results, limitations, tool versions, and recovery paths.
Python · FastAPI · PyTorch · TensorFlow · ONNX Runtime · JavaScript · Bash · SQL · PostgreSQL · Docker · Terraform · AWS · Google Cloud · Cloudflare · GitHub Actions
- Research collaboration: research@secopsai.dev
- Security and vulnerability correspondence: security@secopsai.dev
- Project discussions: SecOpsAI issues
For research or security reports, include reproducible evidence, affected versions or systems, collection dates, and the limits of the available evidence.