Kerollos Ashraf

AI Automation Engineer specializing in production-oriented LLM and RAG systems — spanning intent classification, vector search, and multi-channel AI agents deployed live across WhatsApp, Telegram, and Instagram. Built and shipped five end-to-end AI products independently, from architecture through deployment, for real businesses in hospitality, healthcare, and recruitment. Currently pursuing a B.Sc. in Artificial Intelligence at the Egyptian Russian University.

Seeking a full-time AI Engineering / Automation role to build, test, and optimize production-grade LLM, RAG, and agentic workflows.

AI Automation Engineer at Self-Employed / Portfolio (Present)

Built and deployed multiple production-grade AI and automation systems across different industries.

• Multi-Channel AI Customer Service & RAG System: Designed and built a multimodal intent classifier handling text, voice, and image inputs to automatically route customer queries in real time.
• Engineered a full RAG pipeline — Google Drive watcher, custom text splitter, Google Gemini embeddings, and a Supabase vector store — powering an agent that captures customer data, takes orders, and auto-generates invoices end-to-end.
• Deployed across WhatsApp, Telegram, and Instagram as a single unified customer-facing agent serving ~70–80 customers daily.
• Cut average customer response time from 5–10 minutes to 30–60 seconds and contributed to a ~20% increase in sales.
• AI-Powered HR Recruitment Pipeline: Built a full end-to-end recruitment automation covering application intake, CV screening, written assessment, and phone-screen prep — orchestrated in n8n with multiple chained AI decision points.
• CV Screening: extracts resume text, then an AI agent (Gemini) scores each candidate 0–1 against the job description via structured JSON output, auto-rejecting and logging low scorers.
• Written Assessment & Phone-Screen Prep: dynamically generates and grades 5 role-specific interview questions (50-point rubric), then generates tailored phone-screening questions and emails HR a full candidate summary with both AI scores.
• Included production-grade safeguards: automatic manual-review routing for unreadable resumes, a fallback LLM (OpenRouter), and synced records across Google Sheets, Airtable, and Gmail.
• Clinic Automation Suite (4-part AI system): Secretary Assistant with voice/text chatbot to add, edit, search, and remove patient records, plus send confirmation and reminder emails.
• Patient Assistant: chatbot for booking, canceling, and confirming appointments and general clinic inquiries.
• Reminders & Doctor Assistant: automated day-before reminders confirming required documents (prescriptions, labs, X-rays), a pre-visit patient brief for the doctor, and a note-taking/data-storage assistant during consultations.
• Integrated WhatsApp, Telegram, Google Calendar, and Drive into one workflow; improved patient satisfaction by ~70%, increased on-time attendance, and reduced workload on the secretary and customer service.
• Job Opportunities Aggregator & Application Assistant: Automated real-time tracking of new listings from Mostaql and LinkedIn using n8n workflows and API polling.
• Built an LLM pipeline to summarize postings and generate tailored application drafts, significantly reducing manual search and drafting time.

B.Sc. in Artificial Intelligence – Egyptian Russian University