AI Engineering Studio · Hong Kong

Building Production AI Systems

Retrieval-driven, agentic, and automation-focused AI solutions for modern business workflows.

RAG pipelines · Multi-agent systems · LLM integrations · AI-powered ecommerce · Semantic search

About the Lab

Bahrami Labs is an independent AI engineering studio building retrieval-driven, agentic, and automation systems for real-world business workflows.

Founded by Hossein Bahrami Nekoo, an AI engineer focused on production RAG architectures, multi-agent orchestration, and LLM integrations — with deep roots in ecommerce systems engineering.

Every system built here is designed for real deployment: structured pipelines, observable architectures, and practical AI that solves specific business problems.

RAG ArchitecturesAgentic AILangGraphLLM IntegrationEcommerce AI
15+
Years in production systems
10+
Open-source AI projects
3+
Years in AI/LLM engineering
HK
Based in Hong Kong

Featured Systems

Production-grade AI systems with documented architecture, real deployment, and measurable outcomes.

Agentic RAG System

ClauseGuard

Problem

Legal teams need to scan contracts for risk clauses and answer specific questions — manually, across dozens of documents.

Architecture

Multi-agent LangGraph pipeline: QueryPlanner → Retrieval → RiskAnalysis → Synthesis. Clause-aware chunking with parent-child hierarchy, hybrid embeddings with language-based model routing, Qdrant Cloud for document-scoped vector search.

  • 3 query modes: risk scanning (8 categories), Q&A with page citations, summarization
  • Hybrid embeddings: BGE for English, multilingual E5 for bilingual contracts
  • Deployed: Render (FastAPI backend) + Vercel (Next.js frontend)
PythonLangGraphRAGQdrantHuggingFaceNext.js
AI-Powered Ecommerce

AI Semantic Search for Magento & Adobe Commerce

Problem

Magento's default keyword search fails for natural language queries — users search for 'comfortable running shoes' and get no results.

Architecture

FastAPI microservice generating sentence embeddings via SentenceTransformers, FAISS vector index built from product catalog. Magento module intercepts search requests and queries the AI service, with automatic fallback to native search.

  • ~50ms search latency on CPU for 2,000+ product catalogs
  • Zero Magento core modifications — fully module-based
  • End-to-end pipeline: extraction → embeddings → indexing → API
PythonFastAPISentenceTransformersFAISSMagento Open SourceAdobe CommercePHP
Multi-Agent Framework

SwarmFlow

Problem

Complex tasks require coordinated AI agents — but most frameworks don't provide observable, production-grade orchestration.

Architecture

Leader–worker agent topology built on LangGraph. Leader decomposes tasks and routes to specialized worker agents. Real-time monitoring via FastAPI + WebSockets dashboard. Supports cloud and self-hosted LLMs interchangeably.

  • Leader–worker architecture with task planning and parallel execution
  • Real-time monitoring dashboard with live agent state visibility
  • Plug-in support for Ollama, OpenAI, and any OpenAI-compatible provider
PythonLangGraphLangChainFastAPIWebSocketsDocker

Explore More on GitHub

Additional AI experiments, tools, and open-source projects are available on my GitHub. Explore the full range of work, from production systems to research explorations.

Visit GitHub Profile

Adobe Commerce & Magento Extensions

Production-ready Magento extensions built for ecommerce teams.

AI Content Generator

Generate SEO-optimized product descriptions, meta titles, and meta descriptions in one click — using any OpenAI-compatible AI provider, right inside admin panel.

Save hours on content creation. Choose your AI provider, writing style, and description structure. The extension handles the rest — with full encryption, ACL controls, and PageBuilder support.

Learn More

One-click generation

Generate product descriptions instantly from the admin panel — no page reload.

Preview & approve

Review generated content in a modal before applying to your products.

Multi-provider support

Works with OpenAI, OpenRouter, Groq, DeepSeek, Gemini, or self-hosted endpoints.

Multi-language output

Generate content in any language with full translation and localization.

8 writing styles

Professional, Casual, Luxury, Technical, Playful, Minimalist, Persuasive, Storytelling.

Encrypted API keys

Secure storage using Magento's built-in encryption backend model.

What We Build

Engineering capabilities across the full AI systems stack — from retrieval pipelines to deployed agentic workflows.

Agentic AI Systems

Multi-agent pipelines with autonomous planning, tool use, and decision routing using LangGraph and custom orchestration layers.

LangGraphLangChainOpenAIFastAPI

Retrieval-Augmented Generation

Production RAG systems with hybrid retrieval, smart chunking, reranking, and document-scoped semantic search.

QdrantFAISSSentenceTransformersHuggingFace

AI Workflow Automation

End-to-end automation workflows that replace manual processes with LLM-driven pipelines integrated into existing business systems.

PythonFastAPIDockerWebSockets

LLM Integrations

Connecting any OpenAI-compatible provider — OpenAI, Groq, DeepSeek, Gemini, Ollama — into production applications.

OpenAI APIOllamaOpenRouterGroq

AI Search & Knowledge Systems

Vector-based semantic search replacing keyword matching, with embedding pipelines and real-time query serving at production latency.

FAISSEmbeddingsFastAPIPython

AI-Powered Ecommerce Tools

AI systems purpose-built for ecommerce: semantic product search, AI content generation, catalog automation, and customer intelligence.

Magento Open SourceAdobe CommercePHPPythonLLMs

Multi-Agent Architectures

Leader–worker and peer-to-peer agent topologies for complex task decomposition, parallel execution, and result synthesis.

LangGraphLangChainPythonRedis

Local & Self-Hosted AI

Fully private AI deployments using self-hosted models — no data leaves your infrastructure.

OllamaDockerLinuxHuggingFace

Work with Bahrami Labs

Building a RAG system, agentic workflow, or AI-powered product? Let's talk about the architecture.

We work on production AI integrations, retrieval systems, multi-agent pipelines, and LLM applications — across ecommerce and general business workflows.

What we take on

  • RAG & Retrieval Systems
    Document ingestion, chunking, hybrid search, reranking pipelines.
  • Agentic AI Workflows
    Multi-agent orchestration, tool use, and task automation with LangGraph.
  • LLM Integration
    Connecting any provider into production applications via clean API layers.
  • AI-Powered Ecommerce
    Semantic search, content generation, catalog intelligence for Magento and beyond.
  • AI Architecture Consulting
    System design, pipeline review, and technical direction for AI projects.