DEV Community: Vivek Thakur

I built a local AI assistant fine-tuned on my own conversations

Vivek Thakur — Sat, 24 Oct 2026 07:26:55 +0000

I wanted an AI assistant that is actually mine: it runs on my machine, knows how I talk, and can use my files and shell. So I built JERN, a small open-source assistant around a fine-tuned 1.5B model, and this is how it works and what broke.

Why I built it

Most assistants are someone else's product: they run on someone else's servers, know nothing about my files, and sound the same for everyone. I wanted one that is mine. It should answer in a voice I chose, be able to look at my own folders and run commands on my own machine, and be small enough to understand end to end.

That meant building it myself, with a small open model I could fine-tune on a single GPU.

What JERN is

JERN stands for Just An Experimental Response Network. It has two models and one app:

Chat model: Qwen/Qwen2.5-1.5B-Instruct, fine-tuned on a text file of my own conversations so it answers as JERN and knows how to call tools.
Char model: a GPT-style transformer I wrote from scratch and trained character by character on the same file. It is a learning experiment, not the assistant.
App: a FastAPI server, a single-file web UI with no build step, and one-file-per-tool plugins for files, shell, web search, downloads and system info. Any MCP servers you configure plug in as well.

Everything runs on my machine with one command: python main.py opens the web UI at http://127.0.0.1:8000.

Fine-tuning on my own conversations

The training data is one text file, data/chat/conversations.txt, with conversations I wrote: how JERN introduces itself, how it should talk, and examples of when to call a tool. The parser in model/jern/data.py turns that file into a dataset.

On first run the app downloads the base model and fine-tunes it, which takes 5 - 10 minutes on a GPU and an hour or more on a CPU. The result is saved to out-jern/ and reused on later runs.

Why a 1.5B model? It fits a modest GPU, fine-tunes quickly, and is small enough to read and change the whole pipeline. The trade-off is quality, which I come back to below.

Tools that run on my machine

The part that makes JERN feel like an assistant rather than a chatbot is the tool loop. The model can emit a tool call, the server runs it, and the result goes back to the model for the next step. Each tool is one file under app/tools/, registered in a ToolRegistry, so writing a new one is a small job.

Built-in tools cover files, shell, web search, downloads and system info. On top of that, JERN can talk to MCP servers configured in mcp.json, such as GitHub. The web UI shows a timeline of each tool call (for example, listing a folder) before the final answer appears.

Because the tools are local, JERN can read my real files and run real commands. That is useful and also a risk, so review what it wants to run before you approve it.

The optional Colab mode, and what leaves your machine

My laptop has no GPU, so I added a way to borrow one. The command python main.py colab --gpu=T4 -- serve runs the model on a free Google Colab T4 and connects it to the same local web UI through an SSH tunnel. The tools still run on my machine.

This is not the same as fully local, and it is worth being exact about why. When the model is on Colab, it receives your messages and whatever the tools return: file listings, file contents, command output. Only the execution stays local. If privacy is the point, run the model on your own hardware and skip Colab.

The setup and the tunnel are covered in docs/colab.md in the repo.

Limitations, what's next, and how to try it

JERN is an experiment. A 1.5B model sometimes picks the wrong tool, repeats itself or gets facts wrong, and shell access on a real machine deserves care. I would rather say so than oversell it.

To try it:

git clone https://github.com/IshvaraX/JERN
cd JERN
python -m venv venv
pip install -r requirements.txt
python main.py

The repo is MIT licensed and the docs cover every command, the data format and how to write a tool plugin. Issues and pull requests are welcome, and I would especially like to hear how you handle approving commands that a model wants to run.

Repo: github.com/IshvaraX/JERN

Announcing IshvaraX — A Deep-Tech Startup at the Intersection of Neuroscience and AI

Vivek Thakur — Fri, 02 Jan 2026 15:29:02 +0000

🚀 Announcing IshvaraX

We’re announcing the launch of IshvaraX, a deep-tech startup operating at the intersection of neuroscience, brain-computer interfaces, and predictive artificial intelligence.

IshvaraX is founded on a long-term vision:
decoding human consciousness from neural signals while ensuring advanced AI delivers real public value.

Our Mission

IshvaraX works to reconstruct aspects of human intelligence from brain activity and apply predictive AI to societal-scale challenges.

Our mission spans two aligned fronts:

Advancing neural intelligence through AI trained on EEG and cognitive signals
Deploying AI systems that support essential public services for India

Core Technology

Neural Synthesis AI
AI systems trained on EEG data to map brainwave patterns to intent and action.

Colony Pattern Analytics
Predictive modeling of collective human behavior to anticipate movement, resource demand, and societal trends.

These technologies allow us to reason across both individual cognition and population-level dynamics.

Jan-Seva: AI for Public Utility

Through Jan-Seva, IshvaraX provides free AI-enabled tools to help Indian citizens navigate:

Administrative processes
Legal and regulatory systems
Logistical and informational challenges

This initiative reflects our belief that frontier AI must deliver measurable public benefit.

Neural Ethics & Data Sovereignty

Neural data requires the highest level of responsibility.

At IshvaraX:

All neural data is anonymized and protected
Data governance follows strict sovereignty principles
Core models and datasets remain proprietary and closed-source

Ethics and security are foundational, not optional.

The Website Is Live

Our website is now live and serves as the first public window into IshvaraX—our vision, technologies, and principles.

More technical disclosures and research updates will be shared progressively.

Test post

Vivek Thakur — Sat, 07 Jun 2025 17:08:56 +0000

Test post

🎓 Boost Your Skills with FREE Courses! 🎓

Vivek Thakur — Fri, 20 Sep 2024 08:57:56 +0000

Looking to upskill or dive into new technologies? Check out these FREE courses designed to equip you with essential skills for the future. Whether you’re just starting or looking to add more to your skillset, there’s something for everyone!

💡 Course List:

1️⃣ Postman Student Expert Certification

Master API testing and Postman fundamentals.

2️⃣ HTML & CSS Essentials

Learn to build and style web pages from scratch.

3️⃣ Essentials of JavaScript

Master the fundamentals of the web's most powerful scripting language.

4️⃣ Python Essentials

Kickstart your journey in one of the most popular programming languages.

5️⃣ Bootstrap CSS Essentials

Design responsive websites with ease.

6️⃣ MongoDB Essentials

Learn to work with NoSQL databases.

7️⃣ Figma Essential

Explore the world of UI/UX design with Figma.

8️⃣ Java Essentials

Get up to speed with the core concepts of Java programming.

9️⃣ Essentials of Prompt Engineering

Harness the power of AI with prompt engineering.

🔟 Blockchain Essentials

Get started with blockchain technology and decentralized systems.

1️⃣1️⃣ Cyber Security Essentials

Protect yourself and your systems with foundational cybersecurity knowledge.

1️⃣2️⃣ SQL Essentials

Master database management with SQL.

💥 Why Enroll?

Completely FREE with certification
Flexible learning at your pace
Learn from experts and join a growing community

🔗 Enroll Now:

Don’t miss out! Join today and start your learning journey for FREE!

FreeCourses #Upskill #Learning #CareerGrowth #TechSkills #OnlineCourses

Free UI/UX Development Courses with Certification – Boost Your Skills Today!

Vivek Thakur — Thu, 05 Sep 2024 18:19:02 +0000

Hey everyone!

While browsing the web, I came across some fantastic free courses with certifications that can help you build valuable skills and enhance your portfolio. I'm sharing the links below, and I highly recommend enrolling to take advantage of these opportunities!

Course List [Focused on UI/UX Development]:

Figma 5-Day Bootcamp: Enroll here
HTML & CSS (Beginner-Friendly): Enroll here
Bootstrap: Enroll here
JavaScript (Beginner to Intermediate): Enroll here
Getting Started with Node.js: Enroll here

Do anyone wants to contribute?

Vivek Thakur — Fri, 15 Mar 2024 17:35:16 +0000

Hey guys,

Actually few weeks ago I built my portfolio but the thing is I want to display all my featured post. So Now, I decided to make an API of my own that uses bs4 to scrape my featured post from my profile. Do anyone wants to help?

Link To Github Repo : https://github.com/vivek09thakur/LinkedIn-Scraper-API

Is it just my luck or does the Wi-Fi always drop at the most crucial moments?

Vivek Thakur — Mon, 04 Mar 2024 21:04:04 +0000

Almost every time when I need to Wi-Fi be fast, don't know why it just dropped as hell and I became doomed 😑

Am I the only one who thinks Google Keep should be available for desktops/laptops to??

Vivek Thakur — Sat, 27 Jan 2024 22:51:06 +0000

Hey amazing mens and beautiful ladies,

Your fellow got a problem 😑

As I am used to google keep and I keep using it from 2 years because its easy to use user interface and simple design I am addicted to it also I have many important notes in it but when it comes to windows it sucks there is no version of google keep for desktop!!!!!!!!!!

And now I am stuck, I tried to use Microsoft OneNote and Todo-List but I don't like them.

If somebody knows a solution then comment it down.

View your local React App on your phone.

Vivek Thakur — Mon, 18 Dec 2023 20:15:00 +0000

Have you ever faced a problem while making your site responsive for small mobile device? Yeah chrome web tools were great but sometimes they are not enough. Sometimes we want how it actually looks like on my device.

So here is you do it!

Step 1 :

Make sure your laptop/PC and phone connected to a same network. If you are using your phone's hotspot then it gonna work fine.

Step 2 :

In your terminal run your development server with some aditional arguements.

npm run dev --  --host

Step 3:

Then you should got something like this in your terminal :

Copy the 2nd Network of the terminal and paste it in your phone's browser.

And done! Enjoy the show!!!!!!!!

If the 2nd address doesn't workout try with 1st one.

Thanks for reading

My Github

IMPORTANCE OF GITHUB PROFILE README

Vivek Thakur — Fri, 10 Nov 2023 06:46:15 +0000

Crafting an Impressive GitHub Profile README

In this article, we'll explore how to create an impressive GitHub Profile README to stand out in the developer community.

Getting Started

Creating a New Repository for Your Profile README

Start by creating a new repository with the same name as your GitHub username. This repository will be your profile README repository. For example, if my GitHub username is "vivek09thakur" then the repository should be named "vivek09thakur/vivek09thakur"

Add a README file

While creating this repo you must check "Initialize this repository with: README box"

And make sure that the repo is public

And now you are all set just go to that repo and start editing README.md file of the repo and after you edited and wrote everything about you , just save and commit changes and done.

This will show as a cover page of your github profile.

Just Like this

Some tips from me :D

There are few things which you can do to make your profile readme looks awesome.

Expressing yourself, your passion, your hobbies, things you like, what are you learning and how peoples can reach to you

Describing your languages and tools

For this I just did this in my README file :


**Languages and Tools**

<p align="left">
    <p align="left">
        <!-- Github Stats -->
        <img width="450px" height="auto" align="right" src="https://github-readme-stats.vercel.app/api?username=vivek09thakur&hide_border=True&theme=tokyonight">
        <!-- HTML5 -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/3/38/HTML5_Badge.svg/768px-HTML5_Badge.svg.png?20110131171049' width='38px' height='auto'/>
        <!-- CSS3 -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/6/62/CSS3_logo.svg/638px-CSS3_logo.svg.png' width='38px' height='auto'/>
        <!-- Javascript -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/6/6a/JavaScript-logo.png/900px-JavaScript-logo.png' width='38px' height='auto'/>
        <!-- Bootstrap -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/b/b2/Bootstrap_logo.svg/638px-Bootstrap_logo.svg.png' width='42px' height='auto'height="auto" style="margin-bottom:5px"/>
        <!-- ReactJs -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/a/a7/React-icon.svg/638px-React-icon.svg.png' width='38px' height='auto' style="margin-bottom:7px"/>
        <!-- C++ -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/1/18/ISO_C%2B%2B_Logo.svg/638px-ISO_C%2B%2B_Logo.svg.png' width='36px' height='auto'/>
        <!-- Python3 -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/c/c3/Python-logo-notext.svg/638px-Python-logo-notext.svg.png' width='38px' height='auto' style="margin-bottom:-5px"/>
        <br/>
        <!-- Jupyter Notebook -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/3/38/Jupyter_logo.svg/638px-Jupyter_logo.svg.png' width='38px' height='auto'/>
        <!-- Pandas -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/2/22/Pandas_mark.svg/638px-Pandas_mark.svg.png' width='38px' height='auto'/>
         <!-- colab -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/d/d0/Google_Colaboratory_SVG_Logo.svg/638px-Google_Colaboratory_SVG_Logo.svg.png' width='60px' height='auto' style="margin-left:-10px"/>
        <!-- Tensorflow -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/2/20/Tensorflow-svgrepo-com.svg/638px-Tensorflow-svgrepo-com.svg.png' width='38px' height='auto'/>
        <!-- Pytorch -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/9/99/Pytorch-svgrepo-com.svg/638px-Pytorch-svgrepo-com.svg.png' width='38px' height='auto'/>
        <!-- Django -->
        <img src='https://cdn.worldvectorlogo.com/logos/django.svg' width='30px' height='auto' style="margin-bottom:-5px"/>
        <!-- Flutter -->
        <img src='https://www.svgrepo.com/show/353751/flutter.svg' width='38px' height='auto'/>
        <br/>
        <!-- git -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/3/3f/Git_icon.svg/146px-Git_icon.svg.png?20220905010122' width='38px' height='auto'/>
        <!-- VS code -->
        <img src='https://upload.wikimedia.org/wikipedia/commons/thumb/9/9a/Visual_Studio_Code_1.35_icon.svg/768px-Visual_Studio_Code_1.35_icon.svg.png' width='38px' height='auto'/>
        <!-- andriod studio -->
        <img src="https://upload.wikimedia.org/wikipedia/commons/thumb/9/95/Android_Studio_Icon_3.6.svg/768px-Android_Studio_Icon_3.6.svg.png?20210301045217" width="36px" /> 
        <!-- google idx -->
        <img src='https://idx.dev/images/code-chevron.svg' width='38px' height='auto'/>
        <!-- google cloud -->
        <img src='https://www.svgrepo.com/show/448223/gcp.svg' width='38px' height='auto'/>
        <!-- Firebase -->
        <img src="https://upload.wikimedia.org/wikipedia/commons/4/46/Touchicon-180.png?20200723100028" width="38px" height="38px">
        <!-- Figma -->
        <img src="https://upload.wikimedia.org/wikipedia/commons/thumb/a/ad/Figma-1-logo.png/638px-Figma-1-logo.png" width="40px"> 
        <br/>
</p>

Adding a stat card to your github

This will not only for visitors but also for yourself to get to know how far you are going. To add a stats card you can visit this link and a stats card to your profile.

Still there are many things like your social account links , github trophies but these are some important things from my POV.

Don't forget to checkout my github.

Thanks for reading :D

Logging out !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

My First Cross-Platform Hello-World Flutter App

Vivek Thakur — Mon, 06 Nov 2023 22:02:11 +0000

Feeling so excited and empowered after getting started with Flutter! The possibilities are endless, and I can't wait to create amazing apps with this fantastic framework. 💡📱

Here is the repo of my first hello world flutter app

My Fav VS Code themes in 2023

Vivek Thakur — Sat, 04 Nov 2023 05:50:57 +0000

Exploring Cool Visual Studio Code Themes

Visual Studio Code (VS Code) is one of the most popular code editors in the world, known for its extensibility and vibrant developer community. Among its many features, the ability to customize the editor's appearance through themes is a favorite among developers

1. Dracula Theme

The Dracula Theme is a well-known and widely beloved choice among VS Code users. This theme features a dark, rich background with vibrant and contrasting syntax highlighting. The dark purplish-blue backdrop creates a visually striking environment that is easy on the eyes, making it suitable for long coding sessions. The Dracula Theme is not only aesthetically pleasing but also functional, as it emphasizes code readability and minimizes distractions. This theme has a modern and sleek look that has won the hearts of developers across the globe.

2. Fire Fox Theme

The Fire Fox Theme, inspired by the Mozilla Firefox browser, brings a unique and refreshing style to your coding experience. It is an excellent choice for those who want to create a welcoming and comforting atmosphere while coding.

3. Github Themes

The GitHub Theme is a VS Code theme that closely resembles the default GitHub code viewer. This theme provides a clean and professional look that closely mirrors the code view you would find on GitHub. Its color palette is soft and easy on the eyes, making it ideal for those who want a neutral and unobtrusive coding environment. The GitHub Theme offers a minimalistic and functional approach, emphasizing code clarity without unnecessary frills.

4. Hack The Box

Hack the Box is a visually captivating VS Code theme that takes inspiration from the world of cyberpunk aesthetics. The theme's design features dark and moody colors, reminiscent of the neon lights often seen in cyberpunk settings. The combination of dark backgrounds with bright, neon-like syntax highlighting creates a futuristic and edgy atmosphere. Hack the Box is an excellent choice for those who enjoy immersive and dramatic themes that transport them to a cyberpunk-inspired coding realm.

5. Github Blue

If you're a fan of GitHub and prefer a more vibrant look than the standard GitHub Theme, GitHub Blue may be your ideal choice. This theme retains the clean and professional appearance but introduces a pleasant blue color scheme. The blue hues create a serene and calming ambiance, allowing you to concentrate better while coding. It's a fantastic theme for those who want to combine the familiarity of GitHub with a slightly more dynamic visual style.

Conclusion

Choosing a VS Code theme is a personal preference, and the themes listed here represent a fraction of the diverse options available

Comment your fav VS code theme