aicoding.am
AI AUTOMATION · PROMPT ENGINEERING · VIBE CODING

From Prompt to Production.AI Systems forReal Business Workflows.

I design and build AI-powered workflows, RAG assistants, prompt systems, CRM/ERP integrations and internal tools - with review, constraints and production handoff.

Explore CasesDiscuss a Build
n8nRAGpgvectorOpenAIClaudeCodexBitrix24Acumatica

Public summaries only · Architecture-first · No unsupported metrics · Built for handoff

Automation

Prompts

RAG

MVPs

Integrations

Agents

What I Build

AI systems shaped around business workflows, not tool hype.

Each module starts with the real process, data sources and review boundaries before implementation details are chosen.

module_01

Business workflows

Map manual work, tool handoffs and decision points before deciding where AI belongs.

  • workflow mapping
  • handoff design
  • automation boundaries

module_02

RAG systems

Build retrieval-first assistants that search approved knowledge before generating answers.

  • chunking
  • embeddings
  • vector search
  • answer grounding

module_03

Prompt architecture

Design durable instructions, response formats, guardrails and evaluation checklists.

  • system prompts
  • prompt libraries
  • structured outputs

module_04

AI-assisted development

Use prompt-to-code workflows to ship prototypes, internal tools and UI experiments with review and refactoring.

  • MVPs
  • internal tools
  • code review
  • technical handoff

Operating Protocol

Structured enough for control, lightweight enough for real delivery.

01

Map the workflow

Understand the business workflow, tools, data sources and constraints.

Deliverables: workflow map, tool inventory, constraints list

02

Design the AI layer

Decide where AI is useful: classification, retrieval, generation, routing, extraction or decision support.

Deliverables: AI layer design, prompt strategy, data flow

03

Build the system

Implement workflows, prompts, integrations, vector search, UI and logging.

Deliverables: working prototype, integrations, logs

04

Test against reality

Run edge cases, broken inputs, missing data and hallucination scenarios.

Deliverables: test checklist, edge cases, fix list

05

Deploy and iterate

Ship, monitor, document and improve.

Deliverables: deployment, handoff notes, iteration plan

Declassified Builds

Public summaries of private systems and reusable implementation patterns.

Client names and sensitive data are removed. Architecture, role and stack remain visible.

Private client system / public summary

RAG Knowledge Base Assistant

A retrieval-first assistant pattern for searching approved knowledge chunks before answering operational questions.

Role: Architecture, prompt engineering, workflow automation and implementation planning.

  • Approved source set
  • Text normalization
  • Chunking strategy

Private client system / public summary

n8n AI Automation Pipeline

An automation pipeline pattern that routes incoming work through validation, AI processing, notifications and logging.

Role: Workflow design, n8n implementation, prompt structure and integration mapping.

  • Webhook intake
  • Field validation
  • AI classification

Sanitized client summary

CRM/ERP Integration Workflow

A system pattern for keeping operational records, lead context and notifications aligned across business tools.

Role: Integration planning, API mapping, workflow automation and QA scenarios.

  • Field inventory
  • API capability map
  • Status sync

Stack

Tools are selected around the product, not the other way around.

The stack stays practical: common web tools, clear written context, and verification steps that fit the repository.

AI & LLMs

  • OpenAI
  • Claude
  • Codex-style coding workflows
  • embeddings
  • structured outputs
  • tool calling
  • evaluation prompts

Automation

  • n8n
  • Make
  • webhooks
  • REST APIs
  • cron jobs
  • event-based flows

Data

  • PostgreSQL
  • pgvector
  • Airtable
  • Google Sheets
  • document parsing
  • CSV/JSON pipelines

Web

  • Next.js
  • React
  • TypeScript
  • Tailwind CSS
  • MDX
  • serverless deployment

Business systems

  • Bitrix24
  • amoCRM
  • Acumatica
  • Telegram
  • WhatsApp
  • Chatwoot
  • email systems

Build Log / Field Notes

Technical notes from AI automation, RAG and vibe coding work.

Reusable patterns, constraints and review notes stay close to the work instead of becoming generic AI content.

RAG · published

How to Build a Retrieval-First AI Assistant

A practical outline for assistants that retrieve approved knowledge before generating business answers.

AI Coding · published

Vibe Coding vs Traditional Development

Where AI-assisted development helps, where it needs review and how to keep prototypes production-minded.

Initialize a Build.

Send the workflow, tools involved, current constraint and what a useful first version should do.

AI AutomationPrompt EngineeringRAG SystemVibe Coding / MVP
hello@aicoding.amBack to top