Muhammad Ahmad

AI Product Engineer & Full-Stack Developer · InfiniteGoose

Lahore, Pakistan
ahmadshahbazz@outlook.com

I design and ship AI-powered products end to end, from SaaS platforms and AI pipelines to web apps and browser tools. I work as an Associate Software Engineer at InfiniteGoose, where I build full-stack features and the AI systems behind live SaaS products: LLM workflows, AI video and text-to-speech pipelines, and the deployments that keep them running.

Alongside product work I do research across fields, from AI security and cryptography to computer vision and wireless sensing. My first paper, on running activity recognition on a memory-constrained ESP32, is published in IEEE Access.

I studied Computer Science at the University of Central Punjab (CGPA 3.90/4.00, completed on a 100% merit scholarship). I'm open to AI product and full-stack builds, and to research collaborations.

news

Oct 2026	My first paper is out in IEEE Access (open access): Wi-Fi activity recognition on a memory-constrained microcontroller.
Jul 2026	Graduated with a BS in Computer Science from UCP (CGPA 3.90/4.00) and moved to full-time at InfiniteGoose.
Mar 2026	Joined InfiniteGoose as an Associate Software Engineer.
Feb 2026	2nd place at Pakistan's first Student Project Management Competition, PMI Lahore, pitching SabzChhat Pakistan. Awarded a sponsored PMI delegate seat, South Korea.
Feb 2026	Started as a Research Associate at UCP.
Jan 2026	1st place out of 40 teams at the IEEE × IEC MVP Hackathon with Smart Course Enrollment, which became PlanMyUCP.
Oct 2025	Became President of the PMI Student Club, UCP Branch, and a Teaching Assistant for C++ and OOP labs.

selected publications

IEEE Access
Feasibility and Deployment Analysis of Wi-Fi-Based Human Activity Recognition on a Memory-Constrained Microcontroller

Annas W. Malik, Muhammad Ahmad, and Ki-Il Kim

IEEE Access, 2026 · open access

DOI HTML Code

Which activity-recognition models can actually run on a classic ESP32 (520 kB on-chip RAM, no PSRAM) using only Wi-Fi signals? We trained classical models, tiny CNNs and five deep architectures, quantised them to 8-bit, and measured memory, latency and energy on the board, including a 30-minute live capture-to-prediction run. Memory was not the blocker: the deep CNN and Transformer ran in 10–54 kB of working memory. The harder problem is generalising to new people.
```
@article{malik2026feasibility,
  title   = {Feasibility and Deployment Analysis of Wi-Fi-Based Human Activity
             Recognition on a Memory-Constrained Microcontroller},
  author  = {Malik, Annas W. and Ahmad, Muhammad and Kim, Ki-Il},
  journal = {IEEE Access},
  year    = {2026},
  doi     = {10.1109/ACCESS.2026.3742666}
}
```

projects

Clipsly	Daily AI-made videos for YouTube. Generation and scheduling backend.
SiteVo	Self-building, self-hosting websites. Theming engine.
Vumu	Personal outreach videos with an AI presenter. Pipeline and frontend.
BrainDesk	One support inbox for a SaaS portfolio, with AI-drafted replies.
PlanMyUCP	Course-planning Chrome extension. Hackathon winner.
AnomalyCrypt	Research-based project: face-only encryption with a chaos cipher and a tamper check on arrival.

experience

Mar 2026 – now	Associate Software Engineer, InfiniteGooseFull-stack (React, TypeScript, Vue, FastAPI), LLM workflows with LangGraph and OpenRouter, AI video and text-to-speech pipelines, Docker deployments and on-call.
Feb – Jul 2026	Research Associate, University of Central PunjabResearch that led to the IEEE Access paper.
Oct 2025 – Jun 2026	President, PMI Student Club, UCP Branch
Oct 2025 – Jan 2026	Teaching Assistant, University of Central PunjabC++ and object-oriented programming labs, 40+ students.

education

2022 – 2026	BS Computer Science, University of Central PunjabCGPA 3.90 / 4.00 · completed on a 100% merit scholarship.

skills

Product	React · TypeScript · Vue · FastAPI · Python
AI systems	LangGraph · OpenRouter · LLM workflows · Text-to-speech · AI video
Infrastructure	Docker · deployments · on-call
Research	Machine learning · AI systems · Computer vision · Security · Cryptography · Sensing