How to Clone Yourself with AI in 8 Steps

Why It Matters

This method makes personal digital cloning accessible and affordable, raising practical use cases for personalized assistants while exposing significant privacy, security and ethical risks from training models on private conversations. Organizations and individuals must weigh benefits in productivity and personalization against potential misuse, data leakage and unexpected model behavior.

Summary

A presenter outlines an eight-step process to build a personalized AI “clone” that emulates an individual’s voice and responses. The workflow starts with exporting and cleaning years of Telegram chat, converting conversations into instruction–response pairs, selecting a powerful base LLM, and fine-tuning it efficiently with tools like LoRA/QLoRA and Laura. The model is then trained on the user’s digital footprint, compared against the base model for fidelity, and tested for overfitting, failure modes and odd or privacy-sensitive behaviors. The session promises a live demo showing a full pipeline and results from fine-tuning on five years of chat logs.

Original Description

What happens when you train an AI model on five years of your own conversations?

In this experiment, Alessandro Romano fine-tunes an open-source LLM using years of Telegram chats to see whether AI can learn a person's writing style, humor, recurring expressions, and communication patterns.

The process involves:

✅ Exporting Telegram data

✅ Cleaning and structuring chat logs

✅ Creating instruction datasets

✅ Choosing a base model

✅ Fine-tuning with LoRA/QLoRA

✅ Evaluating whether the model truly sounds like its creator

But the most interesting questions come after training:

Can the AI actually replicate your personality?

How do you measure success?

What are the privacy risks?

And what happens when an AI is trained on your digital footprint?

Learn the complete workflow in Alessandro Romano's session:

🎤 How to Hack Myself: Fine-tuning an LLM on 5 Years of Telegram Chats

📍 DataHack Summit 2026

Register now:

https://www.analyticsvidhya.com/datahacksummit-2026/sessions/how-to-hack-myself-fine-tuning-an-llm-on-5-years-of-telegram-chats?utm_source=social&utm_medium=yt-short

#DataHackSummit #DHS2026 #GenerativeAI #LLM #FineTuning #LoRA #OpenSourceAI