Usama M.

Full Stack AI Developer | RAG, LLM Apps, Document AI | SaaS Products

I build AI products that make it to production: RAG document intelligence, video AI pipelines on self-hosted LLMs, and AI-first SaaS platforms. $200K+ earned, 6,200+ hours, 100% Job Success, Top Rated Plus. WHY FOUNDERS AND TEAMS BRING ME IN: ✅ SaaS platforms, B2B marketplaces, and AI systems shipped since 2020; from idea to production, several still live and earning ✅ Builder of Quanta, and the AI document platform behind 1,000+ SBA loans and $500M+ in funded transactions ✅ Hands-on senior engineer: you work directly with me on architecture and code, on every engagement. When a project needs more hands, I scale with my own vetted team (Arithmiks, 10 engineers) instead of handing you off ✅ Top Rated Plus (top 1% on Upwork). Clear trade-offs, honest assessments, no fluff WHAT I BUILD: 📄 1. AI Document Intelligence and RAG Systems Contract review with clause-level risk flags and exact-section citations. Hybrid retrieval (dense + sparse vectors, Qdrant), structured LLM extraction validated against schemas, audit-logged answers that never cite what wasn't retrieved. Shipped inside a lending platform that has supported 1,000+ SBA loans and $30M+ in funded transactions: AI document naming and summarization, package-level analysis, and credit memorandum generation. 🎥 2. Video and Media AI Pipelines Fully automated pipelines: speech-to-text (Whisper) with speaker diarization, word-level timestamps and subtitles, frame analysis, face detection, chapter segmentation, tag scoring, and natural-language chat over indexed video libraries. Multilingual (including Arabic and English). Deep experience with self-hosted LLM inference (vLLM) on GPU Kubernetes clusters with workflow orchestration, for teams with data-residency or firewall requirements. 💬 3. AI Chatbots and Assistants (with citations, not hallucinations) Chat interfaces over documents, data, and video metadata — every factual answer carries a clickable citation to the source; the system says "not found" instead of guessing. RBAC-scoped retrieval, per-org data isolation, full RAG audit logs. 🚀 4. B2B Marketplaces and FinTech Workflows Vendor portals, quote and order workflows, Stripe payments at real transaction volume, e-signature flows, multi-role approval pipelines, compliance-aware payment architecture. 🏦 5. Cloud, DevOps and Self-Hosted AI Infrastructure AWS (EC2, S3, Lambda, RDS, API Gateway), Docker, CI/CD — and the part most teams can't offer: self-hosted LLM deployments (vLLM, Kubernetes, GPU scheduling) for clients with data-residency or firewall requirements. ☁️ 6. Cloud, DevOps and Self-Hosted AI Infrastructure AWS (EC2, S3, Lambda, RDS, API Gateway), Docker, CI/CD — and the part most teams can't offer: self-hosted LLM deployments (vLLM, Kubernetes, GPU scheduling) for clients with data-residency or firewall requirements. SELECTED PRODUCTS AND CLIENTS: ⭐ Working with Al Jazeera Media Network on AI platform development. ⭐ ClauseLens — AI Contract Review (launching) Upload a 30-page contract, get a complete risk analysis in under 90 seconds: risky clauses flagged red/amber/green with section numbers, key commercial terms extracted, and chat with citations that scroll the original PDF to the exact highlighted clause. ⭐ VirtualBDA / SBA Loans HQ — FinTech Document AI The SBA front-end for lenders: AI document collection, naming and summarization, package-level credit analysis, credit memo generation, e-signatures, live pipeline. Platform stats: 1,000+ SBA loans closed, $500M+ funds. ⭐ Quanta — Conversational Analytics Ask your data anything in natural language — AI chat over your business data. MY STACK: - Python, Django, FastAPI, Node.js / React, Next.js, TypeScript, Tailwind - LangChain, LangGraph, LlamaIndex, CrewAI / OpenAI, Claude, Gemini, Qwen, Whisper - RAG: Qdrant, hybrid dense+sparse retrieval, embeddings, rerankers / vLLM self-hosted inference - PostgreSQL, Redis, MongoDB, MinIO / Kafka, Celery, Argo Workflows - AWS, Docker, Kubernetes, CI/CD / Stripe, DocuSeal, Twilio, AWS Textract I use AI-assisted development tools (Cursor, Claude) to move fast, while keeping full control of architecture and system design. WHY CLIENTS STAY: → 78% of my client relationships run longer than 90 days. → Clients who started with a $100 task have grown into $50K+ engagements. → End-to-end ownership, concept to production — since 2020. If you're building an AI product, need RAG done properly with citations, or have an existing SaaS that needs to scale - send me a message or an invite and tell me what you're working on.