Muhammad J.

I build end-to-end AI SaaS products, from agents, RAG and document intelligence to backend, frontend and cloud deployment. With 5+ years in AI engineering, I combine full-stack product development with LLM fine-tuning, evaluation and inference optimization. I’ve built multi-agentic AI systems across a range of use cases, combining RAG, document intelligence, structured reasoning, tool use, memory, human review, API and MCP integrations, and databases. I also design the full production layer around them, including workflow orchestration, auditability, recovery, reliability controls, and full-stack product engineering. On the model side, I’ve led work across fine-tuning, evaluation, compression, quantization, pruning, multi-GPU execution and optimized self-hosted inference, including leading OptiML, an open-source LLM fine-tuning and compression framework. Top Rated on Upwork · BS Computer Science with honours · 25+ production AI projects · led OptiML model compression and finetuning architecture and releases · two peer-reviewed Medical Image Analysis publications. WHAT I BUILD ✔ AI SAAS & FULL-STACK AI PRODUCTS Production-grade AI applications using Python, FastAPI, React/Next.js and modern data infrastructure, including auth, dashboards, streaming AI, background jobs, integrations, deployment and CI/CD. ✔ AI AGENTS & INTELLIGENCE SYSTEMS Multi-Agentic systems combining LLM reasoning, RAG, memory, structured outputs, business rules and tool use. I connect AI with APIs, databases, CRMs, documents and external tools through function calling and MCP-enabled architectures. ✔ AI INFRASTRUCTURE, FINE-TUNING & LLM OPTIMIZATION Fine-tuning, evaluation and optimization of open-source LLMs using LoRA, quantization, pruning and benchmarking, with multi-GPU execution and optimized self-hosted inference. I also design the production infrastructure around these models for teams that need lower inference cost, stronger privacy, or greater control over deployment. ✔ RAG & KNOWLEDGE SYSTEMS Production RAG over internal knowledge, with citations, persistent context, evaluation, human review, source traceability and governed memory. ✔ DOCUMENT AI & OCR Extraction, classification, validation and reconciliation across PDFs, scans, forms, contracts, invoices and tax documents using OCR, AWS Textract, Azure OCR, vision models, multi-model verification and structured outputs. ✔ BUSINESS INTELLIGENCE & DECISION SUPPORT AI systems that combine business data, deterministic logic and external signals to identify what changed, what needs attention and what action should happen next. ✔ WORKFLOW & OPERATIONS AUTOMATION Intelligent workflows for intake, classification, document processing, reconciliation, approvals, scheduling and follow-up, built to reduce manual handoffs and make multi-step operations faster and more reliable. ✔ VOICE, SALES & CRM AUTOMATION Voice and chat agents for qualification, booking, support and follow-up, integrated with calendars, CRMs and business workflows using LiveKit, Vapi, GoHighLevel, HubSpot, Airtable and n8n. ✔ COMPUTER VISION & EDGE AI Computer vision for pose estimation, gait analysis, activity recognition, facial recognition, image understanding and edge inference using PyTorch, TensorFlow, OpenCV, MediaPipe, MMAction2, MMPose, MMDetection, DeepFace, ONNX, TensorRT, NVIDIA Jetson and ROS. CORE STACK AI & AGENTS LangGraph · LangChain · OpenAI · Claude · Gemini · Grok · OpenRouter · RAG · Tool Calling · MCP · Structured Outputs · Persistent Memory · FAISS · pgvector MODEL ENGINEERING & LLM OPTIMIZATION PyTorch · Hugging Face · LoRA · Quantization · Pruning · Wanda · SparseGPT · PyTorch DDP · lm-eval · SageMaker · Ray · Kubernetes/K3s · TGI COMPUTER VISION & EDGE AI TensorFlow · TensorFlow Lite · OpenCV · MediaPipe · MMAction2 · MMPose · MMDetection · DeepFace · BLIP · ONNX · TensorRT · NVIDIA Jetson · ROS . YOLO BACKEND & DATA Python · FastAPI · Pydantic · SQLAlchemy · Celery · Redis · PostgreSQL · Neo4j · MongoDB · DynamoDB · Supabase · REST APIs AUTOMATION & INTEGRATIONS n8n · GoHighLevel · HubSpot · Airtable · AWS Textract · API Integrations CLOUD, INFRASTRUCTURE & FRONTEND AWS · Azure · SageMaker · Docker · Kubernetes/K3s · Railway · Vercel · GitHub Actions · CI/CD · React · Next.js · TypeScript · Vite · Tailwind HOW I WORK I start with the constraint that matters most: model quality, latency, cost, privacy, workflow reliability or user experience. Then I decide whether the right solution is better prompting/RAG, fine-tuning, model optimization, self-hosted inference, application engineering, or a combination of them. Clear scope. Written updates. Working demos. Documented handover. Building an AI product, agent/RAG system, fine-tuned model, or private/self-hosted AI stack? Send me the problem you’re trying to solve. I’ll help determine the right mix of application architecture, model engineering, optimization and deployment to turn it into reliable production software.