Hi there 😃👋

I'm Amin — nice to meet you!

I’m an AI Engineer at TogetherCrew, where I build data pipelines and LLM-powered systems to support decentralized online communities.

---

💼 What I Work On

• Designing retrieval-augmented generation (RAG) systems using llama-index, with persistent caching, data deduplication, and relevance-based embedding logic.
• Creating ETL pipelines that handle large-scale community data across multiple platforms using Apache Airflow and Temporal. I implement optimizations like indexing only the latest data and deduplicating via hash comparison.
• Evaluating LLM outputs with custom RAG evaluation metrics including coverage, relevance, and confidence scoring.
• Developing LLM agents using LangChain and CrewAI for orchestrated multi-agent tasks.
• Continuously improving data quality and pipeline efficiency by investigating upstream community behavior data and pushing for tighter integration across microservices.

---

🧠 Projects

🔹 Hivemind-Bot — RAG System, LLM

Message-driven system that performs retrieval over embedded organizational data to generate LLM-based responses. Communicates with other services via broker queues.

🔹 Hivemind-ETL — ETL, Caching

ETL DAGs using Apache Airflow for data embedding and summarization. Includes caching mechanisms to avoid redundant embeddings and indexing strategies based on timestamped persistence.

🔹 Temporal Worker — Workflow Orchestration & Data Processing

Implemented scalable, fault-tolerant workflows using Temporal for orchestrating asynchronous data processing tasks. Features include data deduplication via hashing, ETL orchestration, message brokering, and seamless integration with microservices for real-time event-driven pipelines.

🔹 Violation Detection — LLM Classification

Fine-tuned a custom LLM to detect community violations in messages. Built pipelines for classification and automated reporting.

🔹 Agents Workflow — CrewAI, LangChain

Developed multi-agent LLM apps using CrewAI and LangChain, focused on community data analysis and dynamic decision-making tasks.

🔹 TC-Analyzer — Analytics Library

Python library for behavioral analytics on community members. Features graph analytics and activity-based segmentation.

---

🧰 Tech Stack

• Languages: Python, LaTeX
• Databases: Qdrant, PostgreSQL, Neo4j, MongoDB
• Messaging & Workflow: RabbitMQ, Apache Airflow, Temporal
• Frameworks & Tools: Docker, Flask, llama-index, LangChain, CrewAI, Git

---

🙌 Volunteer Work

• Co-Founder, AI Community Group (Nov 2024 – Present) Host a weekly AI series covering topics like LLMs, RAG pipelines, agent systems, and prompt engineering. 🌐 Website

• Co-Founder, Cassandra AI Group (Oct 2021 – Oct 2023) Ran an academic AI community with a focus on accessible ML research and student support. Organized workshops, study sessions, and a two-day conference. 🌐 Website | YouTube | GitHub

---

📫 Get in Touch

• ✉️ Email: dadgaramin96@gmail.com
• 💼 LinkedIn: linkedin.com/in/mramin22
• 💬 Discord: mramin22#1669
• 🐦 Twitter: @mramin22

---

📊 GitHub Stats

---

✨ A Quote to Remember

"After everything, what remains is kindness — so don't hesitate to help others." 😊