Saiam

ApplyPilot helps applicants anywhere in the world apply for jobs, exams, and admissions without losing opportunities to paperwork. Users upload a notice (PDF, photo, or text), and ApplyPilot extracts the key facts, checks eligibility against their profile, answers questions with page-level citations and exact quotes from the notice, screens for scams, fixes photos/signatures to exact portal specifications, auto-fills application forms via a Chrome extension (never submitting — the human always clicks last), and sends deadline reminders (7, 3, and 1 days before, plus overdue) through Telegram and WhatsApp. Meta products are the intelligence layer of this system. The Meta Model API (Muse Spark 1.3) is our default and production model provider, powering notice understanding, structured extraction, cited question-answering, vision-based document reading, and a supervisor-plus-specialist-agents architecture where a supervisor routes each task to a specialist agent with a restricted toolset, behind guardrails. Meta Llama models running locally via Ollama serve as an optional offline backup provider, and local Llama-stack embeddings (nomic-embed-text) power semantic search over notice chunks, because the Meta Model API offers no embeddings endpoint. We also built a WhatsApp Business Platform adapter alongside Telegram, so applicants can use ApplyPilot entirely inside WhatsApp. The whole system — FastAPI backend, Next.js frontend, Chrome extension, and chat bots, containerized with Docker — is verified by 176 automated tests.