AI R&D Engineer · Jakarta
Nadi Rifqi Aufa Rizqullah. Agentic systems, automation backends, and production LLM integrations.
Bridging frontier models with real-world enterprise execution. From prototype to hardened deployment with measurable outcomes.
Selected work
Live systems across HRIS, automation, agents, and generative platforms. Each one proves hands-on delivery.
Full-stack HRIS with digital attendance, leave workflow, AI HR dashboard, Indonesian compliance (BPJS, PPh 21 TER, THR), fraud detection, and real-time SSE notifications.
n8n automation that ingests Gmail, audio, and files, then produces structured Indonesian summaries delivered to Discord.
Pulls news and Facebook signals, ranks trends, and generates Indonesian content briefs with hooks, captions, CTAs, and image prompts.
Event-driven OpenClaw agent that monitors Threads and replies with context-aware LLM responses on a Linux VPS.
Agentic backend PoC that ingests financial metrics and synthesizes actionable forecasting for enterprise workflows.
Scrapes any site DOM, feeds context into an LLM, and outputs a structured design.md master prompt on Cloudflare Pages.
Production batch image generation with public REST API, custom auth, middleware, and rate limiting on a hardened VPS.
Services
From ambiguous AI ideas to production features teams can trust.
Multi-step agents with tools, routing, and recovery paths that survive real traffic.
Secure model wiring with strict schemas, retries, and cost-aware routing.
APIs, queues, and pipelines that turn messy operations into measurable systems.
Auth, rate limits, reverse proxies, and VPS setups ready for production load.
Skills
Credentials
Toolkit
Process
Three non-negotiables before any prompt reaches production.
Unprotected AI endpoints invite billing attacks. Enforce middleware, auth, and rate limiting first.
Never trust raw LLM text. Force Strict JSON Schemas. If it hallucinates, catch and retry.
Move from public APIs to private frontier models on-premise so data never leaves the client VPC.
About
While building AI-assisted workflows for creative products, I got curious about how coding tools manage context, prompts, routing, and multi-step execution. I wanted ideas to move from concept to prototype to deployable product fast.
Instead of treating AI tools as black boxes, I studied workflow patterns: context management, agentic prompt design, request orchestration, and tool-based execution. That improved how I design AI-assisted engineering for creative and automation work.
I applied those learnings into agentic experiments with custom tools including the UpCloud API. The pipeline can provision a Linux VPS, secure a web environment, test generative endpoints, and clean up automatically.
That is how I approach engineering: creative AI experimentation, backend automation, secure deployment, and agentic orchestration turned into working systems.
Contact
Open to collaborations on agentic systems, AI product engineering, and automation infrastructure. Prefer email for concrete briefs.
Browse screens with arrows or thumbnails.