Okafor A.

AI Engineer building production LLM systems, RAG pipelines, AI agents, and AI workflow automation, from architecture to deployment. Python, TypeScript/Node.js, FastAPI, PostgreSQL/Supabase, React/Next.js, REST APIs, and cloud infrastructure. I'm Okafor Chibuzor Authur. I take AI products from concept to production: LLM applications that call tools and APIs, retrieval systems that answer from your own documents, and automated workflows that run reliably at scale without someone watching them. WHAT I BUILD - LLM applications and AI agents: multi-agent orchestration, tool use and function calling, MCP tool servers, structured JSON outputs, and persistent conversation memory. Experience with OpenAI, Claude (Anthropic), and open-source LLMs. - LLM evaluation: evals, test datasets, and error analysis to measure output quality and improve it with evidence, not guesswork. - RAG and document intelligence: document extraction from PDF and Word files, chunking, vector search, and retrieval that gives accurate answers from your own data. - AI workflow automation and integrations: n8n, webhooks, REST API integration, email and SMS workflows, with duplicate prevention, retries, and proper error handling. - Voice and multimodal AI: speech-to-text transcription, text-to-speech, and AI image and video generation pipelines. - Private, self-hosted LLMs: open-source models running on your own infrastructure for data privacy and lower inference cost. - Production backends and full-stack SaaS: Python/FastAPI, TypeScript/Node.js, PostgreSQL/Supabase, SQL, React/Next.js, background jobs, queues and workers, payment integration, webhooks, and deployment to AWS, GCP, Kubernetes, and Linux servers. - Audits and troubleshooting: code review and debugging of existing AI systems, with clear documentation and a practical plan to fix what's broken. SELECTED WORK - Sizemug: Built Weavy, a multimodal AI workspace. An orchestration layer routes each request to text, code, image, video, speech, or tool execution, with an MCP tool server, semantic memory, document ingestion, self-hosted LLMs, and a background job queue with progress tracking and cancellation. - Zenovo AI: Built an AI proposal platform that analyzes 100+ page RFP/RFQ documents with multi-agent LLM workflows, cutting proposal preparation time by 70%. Includes a RAG system on PostgreSQL vector search and a scalable document ingestion pipeline. - Possibl: Architected a multi-agent conversational analytics platform that turns natural-language questions into SQL queries and visualizations, with human approval steps and persistent memory. - Kuuka: Built AI automation for supplier intelligence across thousands of vendor records, integrated financial, analytics, and messaging systems so AI agents can act on business data, and added automated tests across 50+ API endpoints, audit logging, and monitoring. - LearnVault: A multi-tenant AI learning SaaS with transcription, an AI tutor using RAG, and subscription payments with webhook verification. HOW THIS HELPS YOUR BUSINESS - Automate complex workflows that would otherwise need significant manual effort and headcount. - Get accurate answers from your business data instead of time spent searching, verifying, and rewriting. - Handle repetitive, high-volume tasks faster and more consistently with AI agents. - Keep costs in check: I design with LLM cost, reliability, and maintainability in mind from day one. HOW I WORK - Production first: testing, logging, monitoring, retries, and idempotent processing are part of the build, not an afterthought. - AI-assisted development: I use Cursor and Claude Code to ship faster, and I review and test everything they generate before it reaches production. - Independent ownership: I can own the architecture and delivery end to end, or work alongside your existing team. - Clear communication: I explain technical decisions in plain English for founders and non-technical stakeholders, and I document everything I build. - Industries: I've delivered AI solutions across e-commerce, customer support, and internal business automation. - Confidentiality: comfortable working under NDA and strict data-handling requirements. If you need an AI system that works in production, not just in a demo, send me a message with what you're building and I'll tell you how I'd approach it.