Built production RAG systems using Weaviate + LLMs · 3000+ problems solved

Habibur Rahman

$ █

Building production systems with LLMs, RAG pipelines, and modern web tech

View ProjectsDownload Resume
0+
Problems Solved
0+
Years Experience
0+
Projects Shipped
0+
Blog Posts
// About

Who I Am

I'm Habibur Rahman, a full-stack engineer from Bangladesh who builds production AI systems. I specialize in RAG pipelines, LLM integrations, and modern web applications with Next.js and TypeScript.

With 3000+ competitive programming problems solved across Codeforces, LeetCode, and CodeChef, I bring algorithmic rigor to every system I build. Currently working at Makebell, where I architect AI-powered features that serve thousands of users daily.

This portfolio itself is a demonstration — the AI chatbot you see here uses a RAG pipeline I built from scratch, backed by Weaviate vector search and Groq API for near-instant inference.

More about me
~/habib36
$ cat about.json
{
"role": "Full-Stack Engineer"
"focus": "AI & RAG Systems"
"problems": 3000+
"coffee": Infinity
}
// Certifications

Verified Learning

Meta Full-Stack Engineer Professional Certificate

Meta

CourseraJun 2025

Machine Learning Specialization

DeepLearning.AI & Stanford

CourseraDec 2024

LangChain for LLM Application Development

DeepLearning.AI

CourseraSep 2024
// Projects

Featured Work

Production systems I've designed, built, and shipped.

habib36.dev (This Portfolio)

Dark hacker-aesthetic portfolio with an AI chatbot that has read all my content. Built with Next.js 15, Payload CMS, and a custom RAG pipeline.

Next.jsPayload CMSWeaviateTailwind CSSFramer Motion

Natural Language to SQL

AI-powered interface that converts natural language questions into optimized SQL queries. Built with LangChain agents and schema-aware prompt engineering.

92% query accuracySupport for complex JOINsSchema auto-detection
LangChainOpenAIPostgreSQLNext.jsTypeScript

Traffic Signal Detection (YOLOv8)

Computer vision model for real-time traffic signal detection achieving mAP@50 of 0.936. Trained on custom dataset with data augmentation pipeline.

mAP@50: 0.93630 FPS inference12 signal classes
PythonYOLOv8OpenCVPyTorchONNX

Production RAG Pipeline

End-to-end RAG system with Weaviate vector search, LangChain orchestration, and Groq API for near-instant inference. Handles 100+ concurrent queries with sub-200ms latency.

sub-200ms query latency100+ concurrent users95% retrieval accuracy
WeaviateLangChainGroqTypeScriptDocker
View all projects
// Proof of Work

Don't Take My Word For It

Real architecture, real code, real metrics.

System Design

UserNext.js API
│ embed query
APIWeaviate
│ hybrid search
ChunksGroq LLM
│ stream response
LLMUser

Production RAG Pipeline Architecture

Real Codepipeline.ts
█

Measured Results

mAP@50: 0.0

Traffic Signal Detection

<0ms

Vector Search Latency

0%

Retrieval Accuracy

AI-Powered

Ask me anything about Habibur

I've read all his blog posts, project docs, and resume. Try asking me something.

// Blog

Latest Posts

Deep dives, tutorials, and technical opinions.

View all posts

Let's build something together

Looking for a full-stack engineer who actually ships AI systems? I'm open to freelance projects, full-time roles, and interesting collaborations.

Get in touchView resume