depuis la création du compte
Confiez votre projet à Muhammad
Faites appel à l'expertise de Muhammad pour faire avancer votre projet, ou découvrez d'autres freelances pour trouver celui qui correspondra parfaitement à vos besoins.
I'm an AI engineer who builds agentic and multi-agent systems end-to-end: from understanding your business problem to architecture, build, deployment and iteration after launch. I don't just write code; I own the outcome and focus on AI that saves time, cuts costs and drives growth.
What I build:
- AI agents and multi-agent workflows that automate real business processes
- RAG systems and knowledge bases over your documents, websites, videos and data
- Conversational AI and chatbots (web, messaging channels, voice handoff) with human takeover
- Full-stack AI products: React/TypeScript frontends, Python (FastAPI) and Node.js backends on AWS
- Integrations with the tools you already use (Google Drive, OneDrive, SharePoint, Notion, Dropbox) via secure OAuth
Recent experience:
- Forward Deployed Engineer at an AI company: shipping features on a multi-tenant conversational-AI platform orchestrating multiple LLMs (Anthropic, OpenAI, Google), including knowledge graphs (Neo4j), Video RAG and privacy features (PII detection, anonymization, GDPR readiness)
- AI/ML Engineer: built production multi-agent LLM systems for a marketing-intelligence platform and a healthcare assistant, and automated strategy-document generation that replaced weeks of manual analysis; cut LLM Vision costs by about 50%
Stack: Python, TypeScript, React, Node.js, FastAPI, LangChain, LangGraph, LlamaIndex, CrewAI, OpenAI, Anthropic, AWS (Lambda, Bedrock), Docker, PostgreSQL, MongoDB, Neo4j, Redis, Pinecone, FAISS.
Certifications: SAP Certified Generative AI Developer, DeepLearning.AI Deep Learning and Machine Learning Specializations, NVIDIA Deep Learning Institute (Transformers for NLP, Diffusion Models, Deep Learning).
How I work: a short discovery call to understand your goals, a clear scope and plan, regular demos as I build, and clean documented delivery. Tell me the problem, and I'll design the AI that solves it.
Built multi-modal document-ingestion pipelines handling documents, CSV files and Vision-based extraction for enterprise AI products.
- Redesigned the Vision extraction flow into single-pass processing
- Clean structured output ready for RAG and agent workflows
Result: cut LLM Vision API costs by about 50% while keeping extraction quality.
Stack: Python, FastAPI, OpenAI Vision, LlamaIndex.
Automated the onboarding of unstructured PDFs and product manuals into AI workflows, removing manual data preparation for new clients.
- Object- and layout-detection models to understand pages (tables, figures, sections)
- Clean structured output fed directly into downstream LLM workflows and RAG
Result: faster client onboarding and more accurate AI answers from complex documents.
Stack: Python, computer vision / layout detection, LLMs, AWS.
Built and owned key parts of a production conversational healthcare assistant powered by multi-agent LLM workflows and RAG.
- Owned the intent-classification and routing layer, eliminating informational-vs-action misrouting
- Async FastAPI service with thread-safe persistent session management
Result: noticeably better response quality and reliability for end users.
Stack: Python, FastAPI, LlamaIndex, OpenAI.
Competitive, fully funded U.S. Department of State exchange program for undergraduate students.
Skills: leadership, communication, cross-cultural communication, intercultural competence, public speaking, teamwork, networking, adaptability, community service.
Google professional certificate in project management.
Skills: project management, project planning, Agile, Scrum, risk management, stakeholder management, budget management, change management, quality management, Asana.
NVIDIA DLI certificate of competency.
Skills: NLP, Transformers, BERT, large language models, Hugging Face, named entity recognition, text classification, question answering, NVIDIA Triton, deep learning.