01 / SELECTED WORK

Built. Tested.
Still improving.

Projects across models, devices and interfaces. Open a project to see my contribution, decisions and evidence.

Showing 6 of 6 projects.

01Android systems · Community projectSMC CrowsRelease engineering for a community Android game: compatibility, distribution and account recovery.Android release pipelineCommunity engineering · see release evidence
AndroidARM64Release toolingCloudflare

What I built

I work on Android compatibility investigation, release packaging, device testing and the SMC Crows distribution portal. The project also documents installation and account backup and recovery so players can follow the release process.

Result & evidence

The published SMC portal provides release information, setup guides, account recovery guidance and device-test records. Those release-specific records are the reference for compatibility and performance claims.

Illustration · Android release engineering
02Computer vision · CapstoneHybrid Self-CheckoutBarcode scanning paired with computer vision to verify retail items.77.4% precisionReported model benchmark · 50 epochs
YOLOv8OpenCVPythonMySQL

What I built

I built a hybrid validation pipeline that identifies local retail products with YOLOv8 and compares detections with USB barcode scanner input. Python and OpenCV handle the video stream; MySQL stores product and transaction data. Split-view experiments address occlusion by observing products from multiple angles.

Result & evidence

The project repository reports approximately 77.4% precision and 72.0% recall after 50 epochs. These are model evaluation figures, not a measured reduction in retail fraud. The portfolio and resume disagree on inference latency, so no latency figure is presented here.

Illustration · vision + barcode verification
03Agent systems · Edge infrastructureHermesMulti-model orchestration with consensus, telemetry and recovery on mobile hardware.Consensus + recoveryExperimental system · contribution submitted
PythonTermuxLLM orchestrationCloudflare

What I built

The documented Hermes setup runs on a Samsung Galaxy A54 using Termux/Linux. A “Parliament of Minds” routes multiple LLM models through a consensus process for XAUUSD signal experiments. Separate serve and gateway processes, a cron watchdog, telemetry, and causal post-mortems support operation and review.

Result & evidence

The result is a documented multi-model orchestration experiment with observable operations and recovery mechanisms. Historical trading figures on the previous site have not been independently verified and are not presented as evidence of real-world investment performance. The contribution to 500-AI-Agents-Projects is PR #167: open and unmerged when checked on 6 October 2026; that is a historical check, not a current status assertion.

Archived Hermes operations dashboard
Archived dashboard capture · not live telemetry
04Human interaction · Browser MLGesture recognitionA browser-based hand gesture controller for touchless reading.98.6% cross-validationReported score · not real-world accuracy
MediaPipe HandsRandom ForestJavaScriptPython

What I built

IrfanLLM uses MediaPipe hand landmarks and 71-dimensional geometric features. A 100-tree Random Forest trained from video demonstrations is compiled to JavaScript for browser inference. Geometric checks, a 0.30-second scroll cooldown, and virtual air buttons translate gestures into navigation. The existing demo, bookmarklet, and extension remain available.

Result & evidence

The source portfolio reports 98.6% cross-validation accuracy. This is an evaluation on the training workflow’s folds, not a measured accuracy across all users or real-world reading sessions. The working browser controller offers a concrete way to explore that gap.

Irfan demonstrating a hand gesture for the touchless reader
Original gesture training image
05Embedded systems · Team projectIoT livestock monitoringAn Arduino and load-cell weighing system with connected monitoring.INOTEK 2025 · Third PlaceDocumented team award · 15 January 2025
ArduinoHX711Load cellsC/C++

What I built

The resume documents an IoT load-cell weighing platform using Arduino microcontrollers, HX711 amplifiers, C/C++ signal filtering firmware, and real-time telemetry dashboards. I contributed to the weighing and monitoring system as part of the team named on the INOTEK certificate.

Result & evidence

The team received Third Place for “Development of Livestock Weight Tracking System Based on IoT” at INOTEK 2025 (Series 1), held on 15 January 2025. The certificate verifies the team award; it does not validate sensor accuracy or a production deployment.

Illustration · load-cell measurement
06Interactive software · WebMini ArcadeBrowser games exploring decision logic, interaction and application state.Three core games + a betaLive application checked · 6 October 2026
ReactTypeScriptGame logicTailwind CSS

What I built

The original portfolio documents a React and TypeScript arcade with minimax-based Tic Tac Toe, a soccer penalty shootout, and a sports memory match, styled with Tailwind CSS. The current live application lists Pulse Tic Tac Toe, Penalty Shootout, Sports Memory, and the beta Shinobi Legends RPG. Pulse adds rotating rows, decaying marks, and link-based online multiplayer.

Result & evidence

The current application is available as a live demo, with three core games and one RPG marked beta/testing when checked on 6 October 2026. It explores decision-making alongside interface and game-state design. The current game internals were not audited, and no visitor or engagement metrics are claimed.

Mini Arcade interface with its browser games
Original Mini Arcade interface capture

NEXT / A CONVERSATION

Have a role
in mind?

I’m looking for a team where I can contribute, learn and build useful technology.

SELECTED WORK / CASE STUDY

Document preview

Preview the document below, or open the original PDF.

MAKE A CONNECTION

Hello, human.

Complete the verification to reveal direct contact details.

Loading verification…

You can also connect on LinkedIn ↗.