Penetration Tester and AppSec Engineer, building security tools, applied AI products, and systems software in the open.
I help build secure applications by identifying risks before attackers do. By day: 4+ years of VAPT across web, mobile, API, network and thick-client applications, source-code review, cloud security validation and ISO 27001 work at TeamLease RegTech. In the open: the projects below, every one of them working software with tests, not demos.
Latest: RenewalRadar v1.1 (statement-import automation, still fully local) and Cial.ai v1.0 (speak Hindi, send polished English, on Android).
| Project | Domain | What it is |
|---|---|---|
| aegis-security-platform | AI-powered security platform | Aegis: the verification layer for security findings. Ingests Nuclei, ZAP and generic scanner output, deduplicates across tools, verifies each finding with deterministic non-destructive checks, and triages with an AI that advises but never decides the verdict. CLI, API and dashboard; 60 tests including live verification against vulnerable and patched targets |
| adaptive-llm-chatbot | Applied AI / LLM engineering | AdaptiveChat: a document-chat product that sets its own temperature per message, with LangGraph orchestration, input and output guardrails, semantic retrieval over owner-uploaded documents (embeddings with a TF-IDF fallback), LangSmith tracing and a web UI. 54 tests, Docker packaging |
| distributed-task-queue | Systems / distributed systems | A dependency-free distributed task queue in Python: broker with leases and heartbeats, priority queues, retries with backoff, dead-letter queue, at-least-once delivery, persistence, and a worker client. Built to show how systems like Celery and SQS work inside |
| agentforge | AI agents / orchestration | A production-pattern multi-agent framework on LangGraph: supervisor routing across specialist agents, LLM-proposed tool calls validated before execution, a human-approval gate (pause and resume) for sensitive tools, guardrails on input and output, and a complete execution trace per run. 23 tests |
| formfit | Consumer web / public utility | FormFit (live at formfit-lilac.vercel.app): fits a photo, signature or document set to the exact pixel and KB-range rules of online application forms, the rules that reject millions of exam and job uploads every year. Fits both the minimum and the maximum (under-minimum files are padded), converts to JPG, builds PDFs under a size limit, and runs entirely in the browser: zero upload, open source, offline-capable PWA. 8 tests for the fitting core |
| renewalradar | Consumer web / personal finance | RenewalRadar (live at renewalradar-phi.vercel.app): a local-first subscription and bill tracker in dependency-free vanilla JS. Drop in a bank or card CSV statement and it detects recurring charges locally (merchant, typical amount, monthly or yearly rhythm), proposes them for review, and tracks them from there: per-currency monthly burn with currencies never mixed, renewals due in 7 or 30 days flagged, charges projected 90 days out, mark-as-renewed with payment history and price-change flags, and one-click calendar (.ics) reminder export. All data stays in the browser's localStorage, with JSON export and strictly validated import. 73 tests for the date, cycle, detection and totals core |
| cial-ai | Android / AI communication | Cial.ai: chat in any language. A Kotlin Android assistant that floats a microphone bubble over any app when a text field is focused (Accessibility Service, used only for focus detection, overlay and approved insertion). Speak Hindi, Tamil, Telugu, Malayalam, Kannada or English, review the transcript, optionally polish it with your own AI key (OpenRouter, OpenAI or Gemini; English, Hindi or Hinglish output in three tones), and insert it with one tap. No backend, no analytics, the key never leaves the device except to the provider you pick. 33 JVM tests for the prompt, cleanup and settings core; APK on the releases page |
| cloud-misconfiguration-scanner | Cloud security / DevSecOps | Multi-cloud misconfiguration scanner for AWS, Azure, GCP and OCI: public buckets and containers, public policies, and world-open firewall and NSG rules, with offline analysers for exported policies. 27 tests |
| kubeposture | Kubernetes security / Go | KubePosture: a Go CLI that audits Kubernetes manifests against 26 CIS-referenced rules (privileged containers, RBAC wildcards, secrets in env, missing probes and limits, and more), with table, JSON and SARIF output and CI exit codes. 50 tests across loader, rules, reporting and end-to-end layers |
| secretsweep | Secrets detection / Rust | SecretSweep: a Rust CLI that sweeps working trees and git history for leaked credentials: 100 provider patterns (AWS, GitHub, Slack, Stripe, OpenAI and more) plus a Shannon-entropy heuristic, redacted previews and fingerprints only, baselines and allowlists for team adoption, text, JSON and SARIF output. It also opens zip archives, scans their text entries, and warns when an archive pairs an executable with a launcher script, the shape a real malware lure used to hide its payload. v1.2 adds lure and redirector detection: a structural check for remote-config redirector pages, file-host and shortener download-lure links, and instructions to disable antivirus protection. 197 tests including a real git-history integration test |
| pulseboard | Data / TypeScript analytics | PulseBoard: self-hosted, cookieless web analytics in TypeScript. A 693-byte tracker, no raw IPs or user agents stored, unique visitors as a daily-rotating salted hash (labelled an estimate), DuckDB columnar storage, and a server-rendered dashboard. Runs with Docker or plain Node. 61 tests across parsing, privacy, validation, aggregation and HTTP layers |
- OWASP CheatSheetSeries: invited by maintainer @mackowski to review PR #2908 (NPM Security Cheat Sheet, "Inspecting a project you did not write"); review and approval posted, two completeness findings addressed by the author.
- Google osv-scanner: contributor on issue #570 (CLI help documentation).
- pypa/pip-audit: a tested fix for the
--fixside of duplicate requirements (#564), prepared on a fork branch; the upstream repo currently restricts interaction to prior contributors, so submission waits for that to lift.
| Tool | What it does |
|---|---|
| http-security-headers-checker | Checks a site for missing or weak HTTP security headers and scores it out of 100 |
| tls-cert-inspector | Inspects TLS certificates: expiry, issuer, SANs, self-signed and old protocol problems |
| dns-email-auth-checker | Checks SPF, DMARC, DKIM and CAA so a domain cannot be spoofed easily |
| port-scanner-lite | Careful TCP connect scanner with banner grabbing and warnings on risky services |
| jwt-analyzer | Decodes JWTs and audits them: alg none, expiry, sensitive claims, weak HMAC secrets |
| bulk-url-checker | Checks lists of URLs for status, redirect chains and errors, writes a CSV report |
| cors-misconfiguration-checker | Probes a URL with hostile Origin headers and classifies reflection, null-origin and lookalike trust |
| Tool | What it does |
|---|---|
| upload-test-payloads-generator | Generates a suite of harmless file-upload test cases with a manifest of expected results |
| rlo-filename-spoofer | Generates and detects Right-to-Left Override filename spoofs (MITRE ATT&CK T1036.002) |
| pdf-active-content-test-generator | Builds PDFs with benign active-content markers to test upload filters, AV and DLP |
| modern-image-upload-test-corpus | 40 harmless test files for HEIC, HEIF, AVIF, JPEG XL and WebP upload validation and parser robustness |
All tools are for authorised testing, defence and learning. Test files use harmless markers only, never working shells or exploit payloads.
appsec-field-notes: short, practical security posts written from assessment work, covering HTTP headers, file upload testing, JWT audits, email spoofing, filename spoofing, SSRF testing and the IDOR testing method, each with the commands to test it yourself. The latest post dissects a fake-cheat stealer campaign: how lure repos deliver malware, the indicators to spot in two minutes, and how to verify statically without running anything. A new post lands most weeks.
Application and API security, mobile security, network VAPT, OWASP Top 10 and API Security Top 10, SAST and SCA review, LLM application engineering (RAG, agents, guardrails), cloud security across AWS, Azure, GCP and OCI, ISO 27001 and SOC 2 evidence work, TPRM.
- Portfolio: https://rahilsahu.github.io
- LinkedIn: https://www.linkedin.com/in/rahilsahu
- Email: rahilsahu.jobs@gmail.com
- Based in Sagar, Madhya Pradesh, India. Open to remote roles internationally, and relocation for the right role.
Everything published here is for authorised security testing, defence and learning. Only test systems you own or have explicit permission to test.