Syed Muhammad Hassaan

Social links

• 

  

  Global Health

  Lead Software Engineer · Ireland

  Online doctor visits in six countries. I rebuilt the platform and lead it end to end, from architecture and security to automation and deployment, under GDPR and LGPD. 86.5% of bookings need no admin touch.

  • TypeScript
  • PostgreSQL
  • Payments
  • Automation
  • GDPR/LGPD
• 

  

  Endifaa

  Full-Stack Developer · Saudi Arabia

  A sports-academy platform in English and Arabic: user, academy and admin portals, subscriptions and checkout, and full right-to-left Arabic.

  • React
  • TypeScript
  • RTK Query
  • Django/DRF
  • PostgreSQL
  • Docker
• 

  

  Jordy

  Co-Founder · Dubai

  An AI sales rep: it reads your product, finds buyers who need it now, and books meetings through personalised LinkedIn outreach. I built the engine: prospect discovery, research and message writing.

  • React
  • TypeScript
  • LLMs
  • Automation
  • Analytics
• 

  

  Aertonic

  Founding Engineer · Dubai

  Books qualified sales meetings for B2B software companies in the US and UAE, and only charges for the ones that happen. I build the product from zero and set its architecture.

  • TypeScript
  • AI
  • Product Engineering
• 

  

  Zao Fi

  Founding Engineer · Stock AI

  A stock AI that watches what whales, insiders and Congress are buying, and turns it into one 0–100 conviction score with the reasons in plain English.

  • AI
  • Market Data
  • Explainable Scores

• lumen (opens in a new tab)

  Brxerq/lumen

  Something happens on your computer, the devices around you react. An AI coding agent finishes, a build fails or a deploy lands, and your RGB keyboard flashes, your smart lights pulse, your screen edge glows, or you get a notification and a sound. A status tab at the top of the screen shows every open Claude or Codex tab with its activity, context use and usage limits. You set it up in a visual WHEN/THEN rule builder, and it runs on Windows, macOS and Linux with no accounts, no cloud and no telemetry.

  • Python
• claude-cockpit (opens in a new tab)

  Brxerq/claude-cockpit

  Instruments and controls for Claude Code. Live progress bars come from Claude’s own todo list at zero extra tokens, and a router runs quick, normal and hard prompts on the model and effort you pick. A cache meter counts down to the moment the prompt cache expires and can keep it warm. Near the compaction point, Claude chooses to compact the conversation or to write a handoff for a new session.

  • TypeScript
• Super-Market Shelves Detection (opens in a new tab)

  Brxerq/Super-Market_Shelves_Detection_Object-Detection

  The training code, annotated datasets and models behind the shelf-monitoring demo: it finds empty and misaligned spaces on supermarket shelves with SSD MobileNet V2 and transfer learning. Nine models were tested, three SSD MobileNet V2 variants on three datasets (empty, misaligned and both). SSD MobileNet V2 FPNLite 640x640 scored best on empty shelves.

  • Jupyter Notebook

• Shelf monitoring (opens in a new tab)

  Hugging Face Space · Retail computer vision

  Real-time out-of-stock and misalignment detection for supermarket shelves: SSD MobileNet V2 FPNLite with transfer learning and focal loss, trained on 1,000 annotated shelf images. The demo runs three TensorFlow Lite models (both classes, empty only, misalignment only) on CPU with LiteRT. Upload a shelf photo or a video, and it draws a box on each problem and counts them per class.

  • TensorFlow
  • SSD MobileNet V2
  • Gradio
• Face attendance + anti-spoofing (opens in a new tab)

  Hugging Face Space · Multi-layer liveness check

  A liveness check for student attendance: the webcam must show a live person before attendance is marked. It counts 3 slow blinks with dlib face landmarks, checks for one shown hand with MediaPipe, and runs an LBP texture check on the face. YOLOv4 blocks attendance when a phone is in view, such as a photo or video replayed on a phone. The web app saves no images.

  • OpenCV
  • Dlib
  • YOLOv4
  • MediaPipe
• anti_spoofing (opens in a new tab)

  Hugging Face Model · MIT licence

  The real-time face recognition and anti-spoofing system behind the attendance demo. OpenCV, dlib and MediaPipe detect faces, analyse eye blinks and check liveness through texture analysis and gesture detection. YOLOv4 detects smartphones during attendance capture. Made for attendance in educational and professional environments.

  • OpenCV
  • Dlib
  • MediaPipe
  • YOLOv4