AI / RAG AGENTIC AI MLOPS PYTHON LANGCHAIN PYTORCH DOCKER FASTAPI

01 / INTRO

Abdullah Al Arif

AI ENGINEER AI ENGINEER @ FLUVO SOFT

I build AI systems that ship to production — RAG pipelines, agentic workflows, and scalable MLOps. Clarity before complexity.

View Work Contact
Abdullah Al Arif — AI Engineer
2+
Years Exp
In production
65+
GitHub Stars
Open source
4+
Production Systems
Shipped & live
5+
Articles Published
On Medium

About

Turning research into
real-world AI systems.

I'm Abdullah Al Arif, an AI Engineer from Bangladesh. I specialize in architecting production-grade AI systems — from retrieval-augmented generation pipelines to autonomous agentic workflows. My approach is grounded in first principles: I trace the math before trusting the framework, benchmark before optimizing, and evaluate before deploying.

At Fluvo Soft, I'm building agentic AI systems from the ground up — designing multi-agent architectures with LangGraph, integrating MCP for tool access, and structuring workflows for production-level autonomy. Previously at Software Intelligence Ltd, I built RAG pipelines and integrated MCP and A2A protocols end-to-end. That work taught me what most tutorials skip: evaluation, observability, and what happens when your system meets real data.

I believe the engineers who will matter in this decade aren't the ones who know the most APIs — they're the ones who understand the systems deeply enough to fix them when they break. That's the kind of engineer I build myself to be, every day.

Machine Learning

Python PyTorch TensorFlow Scikit-learn Keras XGBoost NLP Computer Vision

Generative AI & LLMs

LangChain LangGraph OpenAI Hugging Face RAG QLoRA FAISS Agentic AI Prompt Engineering MCP A2A CrewAI ChromaDB Ragas LangSmith GraphRAG Hybrid Search LLM Evals

MLOps & Production

FastAPI Docker Kubernetes MLflow ZenML Streamlit Flask AWS SageMaker GitHub Actions

Philosophy

Hard truths about building
AI that ships.

01

Evaluation is the real skill.

Anyone can call an API. The hard part is knowing whether it works — golden datasets, hallucination detection, LLM-as-judge, and drift monitoring are what separate prototypes from systems you can trust in production.

02

Production thinking beats accuracy.

A model that scores 0.95 in a notebook but can't handle latency, cost, or scale is a liability. I optimize for the full stack: serving, observability, failure modes, and the infrastructure that keeps AI reliable under real load.

03

Ship, don't slide.

95% of AI roles are applied, not research. I focus on end-to-end delivery — from data pipeline to deployed API — because companies hire engineers who can build, deploy, scale, and measure. Not just present.

Projects

Selected work that showcases
my skills in AI engineering.

Featured 62

Production RAG Assistant

Production-grade RAG research assistant using hybrid retrieval, cross-encoder reranking, and citation enforcement to deliver grounded answers from research papers.

Python RAG Hybrid Retrieval Cross-Encoder
View Project
Multi-Agent 1

ProjectSmith AI

Turn your startup idea into a validated project plan. Uses a multi-agent workflow to stress-test assumptions, then generates phases, costs, and a downloadable PRD.

Python Multi-Agent LLM
View Project
Deep Learning 3

Reformer from Scratch

A clean PyTorch reimplementation of the Reformer (ICLR 2020), featuring LSH Attention, Reversible Layers, and Chunked FFNs for ultra-long sequences.

PyTorch Transformer LSH Attention
View Project
From Scratch 2

CNN From Scratch

CNN implemented from scratch using NumPy, demonstrating the math and mechanics behind convolutional neural networks. Trained on MNIST achieving ~85% accuracy.

Python NumPy CNN MNIST
View Project
Fine-Tuning

LLM Fine-Tuning Pipeline

QLoRA fine-tuning pipeline for GPT-2 Medium. 4-bit NF4 + LoRA on attention layers, taking you from raw corpus to a generated sci-fi story.

Jupyter QLoRA GPT-2 Fine-Tuning
View Project

Experience

Where I've made an impact.

Fluvo Soft logo
Present

AI Engineer

Fluvo Soft · Part-time

Jun 2026 — Present Remote

Designed and deployed multi-agent architectures with LangGraph, reducing manual workflow processing time by automating task decomposition, parallel execution, and autonomous decision-making across 3+ production pipelines

Integrated MCP (Model Context Protocol) for tool access, enabling agents to interact with 5+ external data sources and APIs with structured tool-call routing and error recovery

Built reliability and evaluation systems including agent behavior tracing, failure recovery protocols, and production-level autonomy scoring — achieving 99.2% workflow completion rate

Software Intelligence Ltd logo
6 Months

AI Engineer Intern

Software Intelligence Ltd · Internship

Dec 2025 — May 2026 Dhaka, Bangladesh · Remote

Built RAG pipelines serving 1000+ internal documents, achieving 92% answer relevance through Pydantic citation enforcement and cross-encoder reranking on production query loads

Developed agentic AI workflows that automated internal data processing, reducing manual operating effort by 40% and shortening task cycle times across 3 core business processes

Integrated MCP and A2A protocols end-to-end, connecting AI agents with 4+ external tools and data sources with structured error handling and retry logic

Managed the full data pipeline from ingestion to LLM response generation, processing 50K+ tokens per query with sub-3-second latency on production hardware

Open Source logo
1 yr 2 mos

Open Source Contributor

Various Projects

Sep 2024 — Oct 2025 Remote

Built and maintained open source training pipelines for models, implementing data preprocessing, fine-tuning, and evaluation stages using Python and PyTorch across 26 repositories

Maintained flagship project Production RAG Assistant (62 stars), developed agentic AI systems with multi-agent orchestration, and merged 27 pull requests across AI open source projects

Open Source

Building in public.
Contributing to the ecosystem.

26
Public Repositories
65+
GitHub Stars
27
Pull Requests Merged

Active maintainer of production-rag-assistant (60★) and reformer-transformer-from-scratch — production-grade AI tools with strong test coverage and clean CI/CD.

Certifications

Validated expertise in
AI engineering.

AI Engineer for Developers Associate

DataCamp

Issued May 2026 · Expires May 2036

Python LangChain RAG LLMs AI Agents Prompt Engineering API Development Vector Databases MCP

Foundation AI and Machine Learning

Microsoft

Issued Feb 2025

Artificial Intelligence Machine Learning Azure AI ML Fundamentals

Blog

Thoughts on AI engineering
and the field.

Get in Touch

Ready to build something
together?

I'm always open to discussing new projects, creative ideas, or opportunities to be part of your vision.

Book a Discovery Call

FAQ

Frequently Asked Questions

What does Abdullah Al Arif do?

Abdullah Al Arif is an AI Engineer who designs, builds, and deploys production-grade AI systems.

  • Retrieval-augmented generation (RAG) pipelines
  • Autonomous agentic workflows with LangGraph
  • Scalable MLOps infrastructure
  • LLM fine-tuning with QLoRA

What technologies does Abdullah specialize in?

Core AI stack: Python, PyTorch, LangChain, LangGraph, OpenAI, Hugging Face.

  • Agentic systems: MCP, A2A protocols, CrewAI
  • Vector search: ChromaDB, FAISS
  • Evaluation: Ragas, LLM-as-judge
  • Deployment: Kubernetes, AWS SageMaker, Docker, FastAPI

What is RAG and how does it improve AI systems?

Retrieval-Augmented Generation (RAG) combines a language model with a document retrieval system.

  • Fetches relevant documents at query time
  • Grounds responses in factual, up-to-date sources
  • Reduces hallucination significantly
  • Eliminates the need for constant retraining

What is Abdullah's experience with production AI systems?

Abdullah has shipped production AI at two companies.

  • Fluvo Soft: Multi-agent architectures with LangGraph, MCP integration, workflow autonomy
  • Software Intelligence Ltd: RAG pipelines, MCP and A2A protocols, full data-to-LLM pipeline

How can I hire Abdullah for AI engineering projects?

Abdullah is available for freelance AI engineering, consulting, and full-time remote roles.

  • Email: aieng.abdullah.arif@gmail.com
  • LinkedIn: Connect directly
  • Contact form on this website
  • Focus areas: RAG systems, agentic AI, MLOps infrastructure

What open source projects has Abdullah contributed to?

26 public GitHub repositories, 65+ stars, 27 merged pull requests.

  • Production RAG Assistant (62 stars) — hybrid retrieval + cross-encoder reranking
  • Reformer Transformer from Scratch
  • CNN from Scratch using NumPy
  • LLM Fine-Tuning Pipeline

What certifications does Abdullah hold?

Two professional certifications in AI engineering.

  • AI Engineer for Developers Associate — DataCamp (May 2026)
  • Foundation AI and Machine Learning — Microsoft (Feb 2025)

What is Abdullah's approach to building AI systems?

First-principles engineering: trace the math before trusting the framework.

  • Benchmark before optimizing
  • Evaluate before deploying
  • Focus on observability and failure modes
  • Build infrastructure for real production load

Does Abdullah work with LLM fine-tuning?

Yes. Hands-on QLoRA fine-tuning with 4-bit NF4 quantization and LoRA adapters.

  • End-to-end fine-tuning pipeline for GPT-2 Medium
  • Raw corpus to generated output
  • LoRA adapters on attention layers

Where can I find Abdullah's blog articles?

Published on Medium at medium.com/@aieng.abdullah.arif.

  • RAG systems and architectures
  • Agentic AI workflows
  • Production ML workflows
  • AI engineering best practices

Contact

Let's start a conversation.

Have a project in mind or want to collaborate? I'd love to hear from you.

AA

Abdullah Al Arif

AI Engineer

Email

aieng.abdullah.arif@gmail.com

LinkedIn

linkedin.com/in/abdullah-al-arif

GitHub

github.com/aieng-abdullah

X (Twitter)

@Austin2882882

Portfolio Assistant

Ask about Abdullah