Static analysis tools flag vulnerabilities line by line. They don't reason about whether an exposed secret in one file, a missing auth check in another, and a permissive CORS policy in a third actually chain together into an exploitable path. That gap is what Kelan Scan is built to close.

Kelan Scan is an agentic repository vulnerability scanner: a five-agent pipeline — Recon → SAST → SCA → Attack-Path → Remediation — where the Attack-Path agent specifically reasons across files to chain findings the way an attacker would, something rule-based static analysis structurally can't do.

• SAST/SCA foundation: Semgrep, detect-secrets, OSV.dev
• LLM verification pass: grounded in a 754-skill cybersecurity library (ATT&CK / NIST / D3FEND mapped)
• Inference: vLLM serving Gemma 3 27B on AMD MI300X via ROCm, with Fireworks AI as fallback

Built for the AMD Developer Hackathon ACT II (Track 3 / Unicorn Track).

Kelan Scan is the second product direction under Kelan Security, my independent security research and engineering project — chosen after evaluating it against an earlier, more research-heavy protocol (below) using a structured Paul Graham–style framework: validated buyer demand, distribution fit, and a technical differentiator competitors can't trivially replicate.

Before Kelan Scan, I spent months building AITP (Adaptive Intent Transport Protocol) — an attempt to embed cryptographic identity, intent declaration, and AI-driven trust evaluation directly into the network transport layer, enforced at the kernel via eBPF/XDP, with post-quantum crypto (ML-KEM-768 + Ed25519). P99 handshake latency: 4.9ms. DDoS resistance: 98.4% packet drop at 0.3% CPU overhead.

It was a serious technical undertaking — full Rust implementation, threat model, deployment architecture, and a completed IEEE-format paper. But when I evaluated it honestly against real buyer demand and go-to-market fit, the answer was clear: compelling research, wrong wedge for a first product. That evaluation is what led directly to Kelan Scan.

IEEE paper (AITP) submitted — sole author, full protocol spec and reference implementation.

Python · Semgrep · OSV.dev · vLLM · Gemma 3 27B · AMD ROCm

Agentic repo vulnerability scanner with cross-file attack-path reasoning. Built for the AMD Developer Hackathon ACT II, Track 3 (Unicorn Track).

Next.js 14 · Node.js/TypeScript · DynamoDB · AWS Cognito · Railway

Migrated from a hackathon-stage Azure demo (SentinelOneX, built for Microsoft Imagine Cup 2026) to a production AWS platform with WebSocket-capable backend hosting and a refined dark UI system (obsidian / signal blue / trust green / threat amber).

Go · Python · FastAPI · OpenTelemetry · Datadog · Vertex AI

Production-grade observability gateway for LLM applications — distributed tracing, latency percentiles, token cost estimation, and prompt injection detection surfaced in Datadog APM. Won the Google AI Catalyst track.

Metasploit Framework — contributed a patch related to CPU processor thread management. Merged by the core team.

Systems        Rust · Go · C/C++ · eBPF/XDP · Linux kernel
Backend        Python · FastAPI · Node.js · TypeScript · Django · Express
Security       Semgrep · detect-secrets · OSV.dev · ATT&CK/NIST/D3FEND
AI/LLM         vLLM · Gemma 3 27B · Claude · GPT-4o-mini · Ollama · Prompt Engineering
Frontend       React · Next.js 14 · TypeScript
Cloud          AWS (DynamoDB, Cognito) · GCP · Azure · Docker · Kubernetes · Railway
Databases      PostgreSQL · MySQL · MongoDB · SQLite · Redis · DynamoDB
Observability  Datadog · OpenTelemetry · Distributed Tracing

AITP: Adaptive Intent Transport Protocol — An AI-Native Network Layer for Secure, Identity-Bound, Intent-Aware Communication in AI Ecosystems and High-Assurance Organizational Networks Submitted to IEEE Transactions on Network and Service Management, 2026 — sole author

IEEE CNS 2026 — separate security research paper, in submission.

• Building Kelan Scan under Kelan Security, my independent security research and engineering project
• Shipping Pramana Security Cloud toward production on AWS
• B.Tech Computer Science Engineering (IoT, Cybersecurity, Blockchain) at Reva University, Bengaluru
• Technical Team Member, Google Developer Groups (GDG) Bengaluru
• Open to remote roles in AI infrastructure, security engineering, and applied AI