I'm Faez Nozir

A FastAPI modular monolith serving document verification, passive liveness, and face matching from a single API. Document requests pass through a quality gate, then a colour + OCR classifier routes each card type to its own parser, with strict or weak validation contracts declared per document. Liveness runs an anti-spoof CNN; face matching runs on deep face embeddings. Hardened with API-key authentication, request-scoped audit logging, and async database access — fully containerised and reproducible from a single Compose file.

An ongoing R&D effort on flagging physically and digitally altered ID documents — started on my own initiative and since taken on by the company. Independent scored checks cover portrait paste seams, ghost-image mismatch, hologram chromaticity, frame geometry, and face-template proportions — each surfaced individually rather than collapsed into a single pass/fail verdict. Built over 30+ measured iterations against a calibrated corpus of genuine and tampered pairs, with rejected hypotheses documented as rigorously as the accepted ones. Whether the accuracy is good enough to be usable is still an open question — measured, not assumed.

The operator-facing console for the verification engine — run document, liveness, and tampering checks from the browser and inspect every individual check score behind a result, not just the verdict. Includes a searchable, sortable transaction log that reconciles engine and vision-model outcomes side by side, plus a full dual-language interface (English and Bahasa Melayu) with the active locale propagated to the backend so API-sourced strings match the selected language.

A JVM bytecode obfuscator built to replace a commercial licence in the SDK distribution pipeline. Rewrites classes through ASM with correct stack-frame recomputation — the genuinely hard part, since malformed frames fail verification on real Android bytecode at runtime rather than at build time. Performs symbol renaming across the class graph, and ships with a local web UI for configuring and running a protection pass.

An existing Android kiosk and counter app for reading national ID cards, owned by a senior colleague — contact-chip reading over serial and USB readers, contactless NFC through both the phone and external readers, live fingerprint verification against the on-card template, and pluggable facial-recognition engines selected by licence configuration. My contribution was revamping the interface and helping lift the NFC, wired-reader and fingerprint flows out of a single monolithic activity.

Developer-facing reference apps published alongside the eKYC SDK, each showing the shortest working integration for a given capability on both platforms. Distributed to Android developers through a Maven repository and kept in lockstep with SDK releases — including migrating consumers onto a second-generation landmark check and gating the discontinued version behind a clear not-supported path.

A native Android eKYC SDK library (v3.0.0) distributed as an AAR — built in Kotlin on top of a third-party biometric SDK. Supports API 26–34. Features include full onboarding, document capture (Custom & Ready-Made UI), passport MRZ scanning, NFC chip reading, passive liveness, smile liveness (active), multi-range liveness (active), and facial landmark detection (v1 & v2).

A native iOS eKYC framework distributed as an XCFramework — built in Swift using a third-party biometric SDK. Modules: Smart Face Capture, Face Auto Capture, Smile Liveness, Multi-Range Liveness, Smart Document Capture, Passport capture, and NFC Passport reading. Built via xcodebuild for both device and simulator targets.

Android demo app (Kotlin) that integrates and tests all 9 features of the eKYC SDK. Supports toggling between Custom UI and Ready-Made UI modes, allowing clients to see both integration styles before adopting the library.

SwiftUI test app for the iOS eKYC framework. Covers Smart Document Digitization, Smart Face Capture, Face Auto Capture, Smile Liveness, Multi-Range Liveness, Passport Capture, Passport NFC, Landmark V1 & V2, Full Onboarding, and Ready-Made UI modes.

Established a DexProtector obfuscation pipeline for both platforms — protecting the Android AAR and iOS XCFramework before distribution. Uses custom XML config rules per platform. Prevents reverse engineering of proprietary biometric algorithms.

A Digital ID prototype built in Flutter — integrates the native eKYC library via platform channels for OCR document scanning, face verification and identity onboarding on Android & iOS. Built for tender presentations to government agencies rather than production release.

Laravel-based admin portal for software license lifecycle management — issue, renew, expire — with user authentication and admin controls. Used to manage distribution of the eKYC SDK to enterprise clients.

A WhatsApp chatbot for a gold dealer — built with whatsapp-web.js and Node.js. Handles sell gold, trade-in gold, and live price check flows via conversational menus. Features a real-time web dashboard with Socket.io for QR auth status, and admin commands to set manual gold prices.

A full luxury personal branding website for a gold dealer in Kota Bharu — built with Express.js and pure HTML/CSS. Features a Malay-language UI with a premium gold aesthetic, multi-page layout (Home, About, Services, Contact), Open Graph meta for social sharing, and structured SEO.

Augmented Reality chemistry education app built in Unity 3D for a local university — visualises Atomic and Ionic Size in the Periodic Table to make abstract concepts tangible for students.

Comprehensive loan management system for a financial services client — handles applications, repayment tracking, customer records, and admin reporting. Delivered as a freelance engagement.

A private medical documentation system built in Laravel. A parent uploads daily wound photos with simple observation fields; a doctor opens a secure per-case QR link to review the healing timeline — read-only, with no account to create. Includes a day-grouped timeline, side-by-side progress comparison, ZIP export of originals, an installable PWA shell, and an admin dashboard with audit logging and soft-delete purge. Deliberately scoped as documentation only — not diagnosis, prediction, or treatment.

A personal eKYC exploration project building a smile liveness detection demo — covering both Flutter (cross-platform) and a native iOS demo app (XCFramework / Xcode). Built to deepen SDK knowledge beyond work scope, testing smile detection, face oval overlay, and liveness pass/fail states.

A Flutter educational game built for my daughter — teaches Malay letters A–Z through interactive drag-and-drop gameplay. Features 26 animal-themed levels, a hint system, scoring, text-to-speech audio, and a child-friendly UI designed for ages 3–7. Supports iOS, Android, and Web.