Senior AI Software Engineer with 4+ years of experience building scalable backend systems and deploying applied AI in production environments, including real-time systems serving tens of thousands of users. Strong background in FastAPI backends, real-time systems, and LLM-powered workflows — covering RAG and agent systems, self-hosted inference, computer vision pipelines, IoT telemetry, and B2B voice AI. Ships in Python (FastAPI, Django) with TypeScript, React, and Next.js when a product needs a full stack. Experienced in cloud-native architectures (AWS, Azure, GCP, Docker, Kubernetes) and CI/CD, with verified scale to 50,000+ concurrent users and Azure deployments serving 100k–1M+ users. Leads backend and AI workstreams for a 3-engineer team and works client-facing from product needs to production outcomes. Research background includes IEEE publications and machine learning for medical imaging in a regulated hospital setting. Motivated to work on high-impact, globally distributed engineering teams and open to relocation across Europe. Languages: English (fluent, C1).
Sweden-based digital transformation company
Lead backend and AI/IoT workstreams for a 3-engineer team: own platform services, agentic features, and production delivery from design through deploy.
Designed and implemented OptySleep recommendation engine with Sentence-BERT semantic similarity, Redis caching, and A/B testing — ~70% lift in iOS engagement and improved retention and subscription renewals.
Delivered production-grade IoT backend for Staging Pro using Django, PostgreSQL, and Redis, processing live telemetry with <100ms latency for real-time race coordination.
Built PseudoCompiler (FastAPI, GPT-4 + LangChain, WebSockets, Docker, AWS) scaling to 50,000+ concurrent users with <50ms latency; reduced AWS infrastructure cost by ~35%.
Build Voca, a FastAPI B2B outbound voice AI platform with Retell AI, CRM webhooks, lead scoring, and campaign orchestration at ~$0.10 USD per minute.
Introduced containerized deployments (Docker, Kubernetes) and automated ops workflows with RAG + n8n, saving ~8 hours/week.
Prototyped lightweight, quantized LLMs (PyTorch, Ollama) for secure on-device inference on IoT hardware; currently evaluating FunctionGemma for agentic workflows.
Contribute to product work including CarBNB — a car rental app for Gulf and South Asian markets — alongside core AI platform delivery.
Rapid AI software services company
Architected hybrid systems (monolithic + microservices) on Azure with Kubernetes, scaling reliably to 100k-1M+ users.
Shipped AI Legal Assistant (Next.js UI + LangChain RAG, Gemini, MongoDB vector embeddings); citation-backed Q&A cut document review time by ~70%.
Built ViralMe AI Video Editor (viralme.today), using Remotion, OpenCV, FFMPEG, and Node.js rendering workers — reducing turnaround from hours to minutes.
Deployed self-hosted LLMs and Whisper3 via Ollama CLI, optimized GPU usage for cheaper inference.
Developed multi-agent video pipelines (YOLOv12, DeepSORT, SAM2 + RabbitMQ orchestration + LLM agents) to enable async detection, tracking and semantic summarization — achieved near-real-time throughput for production feeds (~20–25 FPS on target infra) and automated tagging.
Automated workflows and built connectors and background workers that cut manual reconciliation and data sync tasks by ~50%, improving reliability and reporting accuracy projects like Fortify ERP (fortify.biz).
Designed REST APIs & microservices (FastAPI/Django, Postgres, Redis, Firebase) and improved reliability & throughput.
Implemented CI/CD (GitHub Actions + Docker), shortened release cycles ~40%.
Migrated services to cloud infra; built dashboards improving issue detection by ~30%.
Digital Health Research Centre · leading academic hospital
Built ML pipelines to support AI-powered breast cancer diagnosis on mammography and ultrasound, achieving 92% accuracy and reducing radiology review time by 87% in a regulated hospital setting.
Worked in a lab incubation environment tied to FYP research (Mammory) in collaboration with Aga Khan University Hospital.
Developed tools for automated triage, sorting, and tagging of radiology images.
COVID-19 audio ML research organisation
Prepared and analyzed large volumes of breath & cough data for early COVID-19 detection using signal processing.
Achieved accuracy of 89% in early-built prototype.
Developed a prototype of a COVID-19 detection solution using Flask & Ensembling techniques, under the supervision of Stanford University & DetectCovid.
My experience as a Microsoft Student Ambassador has been incredibly rewarding, simultaneously supporting two communities. I relish being involved with both GDSC and MLSA, navigating various technologies with my peers. It's an entirely new kind of adventure for me.
Co-established the student developers community in 2019 at our campus. Led the community 2020 to 2021. Hosted total 150 workshops with 100 as self-trainer in AI. Trained over 1000+ students myself alone.
Production-grade IoT backend system for real-time race event coordination. Processes live telemetry with <100ms latency, enabling race coordinators to manage concurrent events, control staging sequences, and broadcast safety flags to connected devices.
Built OptyAI—a custom recommendation engine from scratch that powers personalized sleep insights. Achieved ~70% boost in engagement, improved retention, and higher subscription renewals through Redis caching, A/B testing, and semantic similarity techniques.
Voca is a B2B outbound AI voice calling platform for campaign creation, lead filtration, lead management, and automated cold calling at $0.10/min — cutting large-scale outreach costs for software, pharma, insurance, and debt collection teams. Multilingual voice agents with CRM-ready workflows.
Real-time video editor on the web using Remotion. Edit, preview, and render videos directly in the browser with AI-informed effects. Features include 2-minute transitions from long videos to multiple reels, motion graphics, background noise removal, and more.
Online platform for real-time AQA Pseudo to Python 3 translation with live execution and output visualization. Features complex lexical code translation with integrated AI chat for error detection and code assistance.
Built a Gemini-powered Legal AI Assistant with RAG (Retrieval-Augmented Generation) capabilities. Provides intelligent legal document analysis, question answering, and document retrieval using advanced language models and vector search.
Forever free background remover powered by U2Net deep learning model. Remove backgrounds from images instantly with a simple PWA interface connected to a FastAPI backend.
Natural Language to SQL compiler enabling non-technical teams to query databases instantly. Enforces strict schema validation and structural output guarantees to prevent hallucinations and SQL injection risks.
Selected peer‑reviewed work, conference papers, a published book, and thesis with direct DOI and publisher links.
M. H. Shahbaz, Nadeem Ullah, M. Ahmed
Spectrum of Engineering Sciences ISSN (e)3007-3138 (p)3007-312X vol. 4, no. 5, pp 1724–1734, 2026
Ali, U., Kandhro, I.A., Ahmed R.S., Khan, A.A. Shahbaz, M.H. Osama, M.
International Journal of Emerging Sciences and Digital Economy (IJESDF), vol. 17, no. 3, pp 391-403, April 2025
K. Mahboob, M. H. Shahbaz, F. Ali, and R. Qamar
VFAST Transactions on Software Engineering, vol. 11, no. 2, pp. 249–255, Jun. 2023
M. H. Shahbaz, Zain-Ul-Abidin, K. Mahboob and F. Ali
2023 7th International Multi-Topic ICT Conference (IMTIC), Jamshoro, Pakistan, 2023, pp. 1-7
T. Mubeen, Zain-Ul-Abidin, M. H. Shahbaz, P. O. Roth and M. A. L. Nieto
2023 Global Conference on Wireless and Optical Technologies (GCWOT), Malaga, Spain, 2023, pp. 1-7
Microsoft for Startups · Founders Hub (Transpify)
Awarded $25,000 in Azure cloud credits through Microsoft for Startups Founders Hub to accelerate Transpify infrastructure, AI workloads, and production deployments.
Sir Syed University of Engineering & Technology (SSUET)
Won first place in the university-wide final year project competition for Mammory — AI-assisted breast cancer detection on mammography and ultrasonography.
Google Developer Student Clubs & Microsoft Learn Student Ambassadors
Delivered workshops and mentorship across AI, machine learning, and cloud topics—personally training 1,000+ students through GDSC and Microsoft Learn Student Ambassador programs.
Microsoft
Issued Jan 2021
Credential ID wnYYD-48DY
Microsoft
Issued Dec 2020
Credential ID wnqmz-48Eq
IBM
Issued May 2020
Credential ID MBJSPDSFEMQA
IBM
Issued May 2020
Credential ID 4A2F9HRJJC9G
