I ship production AI systems (LangGraph agents, RAG, on-device YOLOv8) and Flutter apps with Clean Architecture. 4 projects live. MS CS in progress. Based in Peshawar, working worldwide.
I use the OpenCode CLI and AI agents via OpenRouter to speed up planning, building and review.
The same tooling that speeds up my build pipeline — available to you here.
AI systems and mobile apps for startups and product teams.
LangGraph multi-agent pipelines, RAG with evaluation harnesses, prompt guardrails, cost/latency budgets.
Halal-Guard: Flutter + YOLOv8 detecting non-halal ingredients in food images. 22ms inference on Snapdragon 8 Gen 2. ONNX runtime, Firebase sync, full PRD/SRS documented.
Halal-Guard: YOLOv8 on-device at 22ms. ONNX runtime, quantization, cloud fallback patterns.
Halal-Guard: Flutter + YOLOv8 detecting non-halal ingredients in food images. 22ms inference on Snapdragon 8 Gen 2. ONNX runtime, Firebase sync, full PRD/SRS documented.
Task Manager: Flutter + Dart, auth + real-time sync, 60 FPS. PRD/SRS pre-implementation.
Task Manager: cross-platform Flutter task manager with auth and real-time sync. 60 FPS fluid animations. PRD/SRS documented before implementation. Live API integration.
Prediction Agent: multi-source ingestion > ML confidence scoring > microservice with graceful degradation.
Intelligent Data Prediction Agent: multi-source data ingestion, ML confidence scoring, REST API microservice with graceful degradation. Python, web scraping, data pipelines.
Every project starts with PRD/SRS/FSD. Reduces rework, aligns stakeholders, creates audit trail.
Spec-driven by default: PRD/SRS/FSD documented before code. Clean Architecture (data/domain/presentation), BLoC/Riverpod, comprehensive docs enable bus-factor mitigation.
IoT Energy Platform: device-agnostic hardware, AI copilot insights, offline-first sync, smart billing.
Smart Energy Management Platform: PRD/SRS for device-agnostic IoT. AI copilot layer for real-time insights + next-action recommendations. Offline-first sync, smart utility billing. Flask/Django, REST APIs, Docker, K8s.
Selected AI, computer vision and mobile projects — from specification to working product.
Building AI-powered mobile applications with computer vision capabilities, integrating machine learning models for image recognition and intelligent data processing.
Designed PRD, user story, and SRS for a cross-platform IoT application. Specified an AI copilot layer providing real-time system insights and next-action recommendations, with device-agnostic hardware integration and a smart utility billing engine.
Developing AI agent systems for data processing, prediction, and intelligent automation using machine learning models, exposed as clean backend microservices.
Built a cross-platform mobile application end-to-end following professional engineering standards, with full PRD and SRS documented prior to implementation. Includes live API integration and responsive design.
Fork of stickerdaniel/linkedin-mcp-server extended with typed write/action MCP tools for agent-driven LinkedIn control: posting, engagement, connections, job actions, follow, notifications, settings.
Technologies I work with across AI, mobile and backend development.
Work experience and education.
Professional resume showcasing AI engineering expertise, mobile app development skills, and software architecture experience.
Advanced specialization in Artificial Intelligence and Machine Learning, Multi-Agent Systems, and Distributed Systems. Researching high-throughput vector retrieval RAG pipelines, on-device mobile AI compute engines, and real-time speech agent systems.
Developed and maintained features for Flutter mobile applications during internship. Engineered responsive widget components, state management, and seamless cross-platform performance across Android & iOS devices.
Delivered full-stack and mobile development projects for freelance clients. Owned architecture and development end-to-end, integrating REST APIs, database schemas, and AI workflow components.
Graduate specializing at the intersection of AI engineering and mobile app development. Experienced in building AI-powered applications and delivering production-quality software end to end.
Completed Higher Secondary School Certificate in Pre-Engineering, establishing analytical mastery in advanced mathematics, physics, and computational logic.
Secondary School Certificate with foundational honors in science, mathematics, and computer sciences.
Have an AI or mobile project in mind? I would be glad to talk it through. I am open to full-time roles, contract work and consulting.
Who I am, what I build, the stack I use, and how to work with me.
Muhammad Roman is an AI systems architect and cross-platform mobile engineer based in Peshawar, Pakistan. He builds agentic AI systems, LLM-powered applications, computer-vision mobile apps and scalable backend services, and is pursuing an MS in Computer Science.
Muhammad Roman specializes in AI engineering and mobile development: agentic AI and multi-agent systems, LLM integration with RAG and vector search, computer vision (YOLOv8/ONNX), and production Flutter/Dart apps built with Clean Architecture and offline-first logic.
Muhammad Roman works with Python (FastAPI, Django, LangChain), Flutter and Dart, TypeScript and React, Node.js, PostgreSQL, MySQL, MongoDB, Redis, vector databases, Docker, Kubernetes and cloud platforms including AWS, Azure and GCP.
Yes. Muhammad Roman is open to freelance, remote and full-time work in AI engineering, mobile development and full-stack roles. The fastest way to reach him is by email at romankhan7475@gmail.com or through LinkedIn.
Muhammad Roman built Halal-Guard, a Flutter and YOLOv8 app that flags non-halal ingredients; world-news-channel, a LangChain multi-agent news summarizer; a cross-platform Flutter task manager with real-time sync; and several TypeScript and Python production projects.
Muhammad Roman is based in Peshawar, Pakistan, and works remotely with clients worldwide. The best way to contact him is by email at romankhan7475@gmail.com, or by message on LinkedIn.