DEV Community: Sakshi Yadav

Bulk Receipt Sorter: Automated Receipt Processing & Excel/CSV Export App

Sakshi Yadav — Wed, 21 Oct 2026 07:42:44 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built
Sloths have the right idea: why do manual work when you can automate it? For this week's theme, I upgraded my local receipt extractor into a Bulk Receipt Sorter.

Instead of uploading invoices one by one, freelancers and small business owners can now upload 10+ receipts at once. My app extracts the Vendor Name, Date, and Amount from all of them and compiles them into a single CSV file (which opens in Excel) ready for their accountant. This truly "rewards laziness" by saving hours of manual data entry.

Demo

Code
Since this is a local Python app running via Streamlit, here is the core logic that handles the bulk processing and CSV generation:

python

Core loop for bulk processing
for i, uploaded_file in enumerate(uploaded_files):
image = Image.open(uploaded_file)
extracted_text = pytesseract.image_to_string(image)

prompt = f"Extract Vendor, Date, Total Amount from: {extracted_text}"

response = ollama.chat(model='llama3', messages=[{'role': 'user', 'content': prompt}])

Append to dataframe...

How I Built It
· UI: Streamlit (Python)
· OCR: Tesseract OCR (to read text from images)
· AI Engine: Ollama running Llama 3 locally. No cloud APIs were used.
· Data handling: Pandas (for creating the downloadable CSV).

The app loops through all uploaded images, extracts the raw text, asks the local Llama 3 model to structure it, and builds a downloadable table.

Why Does Open Innovation Matter?
Because this process involves hundreds of receipts over a month, using a paid cloud API (like OpenAI) would cost a lot of money and take a long time due to network requests.
By using Ollama to run Llama 3 locally:

It's 100% Free: I can process 500 receipts and it won't cost a single rupee in API fees. No token limits.
Privacy: Freelancers don't want to send 50 images of their client bills to a cloud server. This runs completely offline on the laptop.
Speed: No internet lag. The AI processes the images directly on the local CPU.
The Lazy Payoff
This app saves a freelancer hours of tedious data entry. Instead of opening 50 PDFs and typing numbers into Excel, they can dump all 50 images into this app, click one button, and download a ready-to-use Excel sheet.

Prize Categories
I am entering the Open Source AI category.

Bulk Receipt Sorter: Automated Receipt Processing & Excel/CSV Export App

Sakshi Yadav — Tue, 20 Oct 2026 14:57:25 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady

What I Built

Sloths have the right idea: why do manual work when you can automate it? For this week's theme, I upgraded my local receipt extractor into a Bulk Receipt Sorter.

Demo

Code

Since this is a local Python app running via Streamlit, here is the core logic that handles the bulk processing and CSV generation:

python

Core loop for bulk processing

for i, uploaded_file in enumerate(uploaded_files):
image = Image.open(uploaded_file)
extracted_text = pytesseract.image_to_string(image)

prompt = f"Extract Vendor, Date, Total Amount from: {extracted_text}"

response = ollama.chat(model='llama3', messages=[{'role': 'user', 'content': prompt}])

  
  
  Append to dataframe...

How I Built It

· UI: Streamlit (Python)
· OCR: Tesseract OCR (to read text from images)
· AI Engine: Ollama running Llama 3 locally. No cloud APIs were used.
· Data handling: Pandas (for creating the downloadable CSV).

The app loops through all uploaded images, extracts the raw text, asks the local Llama 3 model to structure it, and builds a downloadable table.

Why Does Open Innovation Matter?

Because this process involves hundreds of receipts over a month, using a paid cloud API (like OpenAI) would cost a lot of money and take a long time due to network requests.
By using Ollama to run Llama 3 locally:

It's 100% Free: I can process 500 receipts and it won't cost a single rupee in API fees. No token limits.
Privacy: Freelancers don't want to send 50 images of their client bills to a cloud server. This runs completely offline on the laptop.
Speed: No internet lag. The AI processes the images directly on the local CPU.

The Lazy Payoff

This app saves a freelancer hours of tedious data entry. Instead of opening 50 PDFs and typing numbers into Excel, they can dump all 50 images into this app, click one button, and download a ready-to-use Excel sheet.

Prize Categories

I am entering the Open Source AI category.

Side Hustle: A Private, Open-Source AI Copywriter for E-commerce Sellers

Sakshi Yadav — Sun, 18 Oct 2026 16:12:05 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built
I built a simple, local AI product description generator aimed at independent e-commerce sellers (like Etsy or eBay shop owners).

The problem is that small sellers often spend 15-20 minutes writing a single listing because they aren't professional copywriters. Closed-source AI tools (like ChatGPT) require a subscription, and sellers might feel uncomfortable uploading unreleased product photos or details to a big tech server. My tool solves this by taking a few plain-text features and instantly generating polished, multiple-option descriptions entirely on the seller's own laptop. It is small, specific, and solves a real time-consuming problem for freelancers and small business owners.

Demo
Here is a look at the tool running locally via the command line. In this screenshot, you can see the Python script being executed in Windows PowerShell. The user inputs the product name "Handmade Silver Bracelet" and a few key features (turquoise stone, adjustable rope, bohemian style). The local AI model then generates three distinct description options with different tones (Warm, Concise, and Detail-focused) right in the terminal, proving the tool runs 100% offline and privately.

https://i.imgur.com/your-image-link-here.png

Note: The image above shows the exact terminal output of the model generating three options and asking follow-up questions to tailor the description further.

Code
I wrote a simple Python script that uses the local Ollama library. Here is the core logic:

python
from ollama import Client

Create a client that ignores system proxies to ensure local connection works

client = Client(host='http://127.0.0.1:11434', trust_env=False)

def generate_description(product_name, features):
response = client.chat(
model='gemma3:1b',
messages=[
{'role': 'system', 'content': 'You are a helpful assistant that writes short product descriptions for small online sellers.'},
{'role': 'user', 'content': f'Product: {product_name}. Features: {features}. Write a short description.'}
]
)
return response['message']['content']

product = input("Product name: ")
features = input("Features: ")
print("\nGenerated description:\n")
print(generate_description(product, features))
How I Built It
I built this entirely around local inference using an open-weight model. I installed Ollama on my Windows PC and pulled the gemma3:1b model (a 1-billion parameter model that runs easily on a standard CPU laptop). I then used Python and the ollama library to connect to the local Ollama server (running on port 11434). The open-source agent harness (Ollama) is what makes the whole thing work without relying on the internet or external APIs.

Why Does Open Innovation Matter?
Open innovation was crucial for this project for two main reasons:

Data Privacy: Etsy sellers often work with custom, unreleased designs. Because the open-weight model runs 100% locally, their product data never leaves their machine.

Zero Cost: Small business owners have tight margins. Using a closed API would cost money per token. By running an open-weight model, the seller can generate hundreds of descriptions for free.

No Internet Required: Once the model is downloaded, the tool works perfectly offline, making it reliable for sellers in areas with poor internet connectivity. It also avoids getting locked into a specific vendor's pricing.

My First User
I tested this with a friend who runs a small handmade jewelry shop. They gave me the "Handmade Silver Bracelet" details. Their reaction was: "Wow, it actually sounds like me, but better. I usually just write 'nice bracelet' and give up." I asked if they would pay $5 for a simple desktop app that did this for all their listings. They said yes, but they asked if the next version could also generate Instagram hashtags automatically. This feedback showed me that the tool has real value and that there is a clear path to adding more features for niche sellers.

Side Hustle: A Private, Open-Source AI Copywriter for E-commerce Sellers

Sakshi Yadav — Sat, 17 Oct 2026 08:40:08 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built
I built a simple, local AI product description generator aimed at independent e-commerce sellers (like Etsy or eBay shop owners).

Code
I wrote a simple Python script that uses the local Ollama library. Here is the core logic:

python
from ollama import Client

Create a client that ignores system proxies to ensure local connection works
client = Client(host='http://127.0.0.1:11434', trust_env=False)

product = input("Product name: ")
features = input("Features: ")
print("\nGenerated description:\n")
print(generate_description(product, features))

How I Built It
I built this entirely around local inference using an open-weight model. I installed Ollama on my Windows PC and pulled the gemma3:1b model (a 1-billion parameter model that runs easily on a standard CPU laptop). I then used Python and the ollama library to connect to the local Ollama server (running on port 11434). The open-source agent harness (Ollama) is what makes the whole thing work without relying on the internet or external APIs.

Why Does Open Innovation Matter?
Open innovation was crucial for this project for two main reasons:

Data Privacy: Etsy sellers often work with custom, unreleased designs. Because the open-weight model runs 100% locally, their product data never leaves their machine.

Zero Cost: Small business owners have tight margins. Using a closed API would cost money per token. By running an open-weight model, the seller can generate hundreds of descriptions for free.

Side Hustle: A Private, Open-Source AI Copywriter for E-commerce Sellers

Sakshi Yadav — Sat, 17 Oct 2026 06:23:13 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

What I Built
I built a simple, local AI product description generator aimed at independent e-commerce sellers (like Etsy or eBay shop owners).

Code
I wrote a simple Python script that uses the local Ollama library. Here is the core logic:

python
from ollama import Client

Create a client that ignores system proxies to ensure local connection works

client = Client(host='http://127.0.0.1:11434', trust_env=False)

product = input("Product name: ")
features = input("Features: ")
print("\nGenerated description:\n")
print(generate_description(product, features))

Why Does Open Innovation Matter?
Open innovation was crucial for this project for two main reasons:

Data Privacy: Etsy sellers often work with custom, unreleased designs. Because the open-weight model runs 100% locally, their product data never leaves their machine.

Zero Cost: Small business owners have tight margins. Using a closed API would cost money per token. By running an open-weight model, the seller can generate hundreds of descriptions for free.

My Hacktoberfest Journey: Building an AI that touches grass so I don't have to

Sakshi Yadav — Sun, 11 Oct 2026 15:05:08 +0000

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

I built an offline, open-source AI nature identifier. It is a Python CLI tool that runs the LLaVA vision model locally on your laptop using Ollama.

The purpose of this tool is to get people outside and exploring nature without the need for an internet connection. It allows hikers, birdwatchers, and gardeners to identify plants, birds, or animals right on their machine, ensuring absolute privacy and zero reliance on the cloud.

Code

You can view all the code, the bird picture, and the log file in my GitHub repository here:
https://github.com/sakshi-yadav-06/touch-grass-ai

How I Built It

I built this using open-source, local-first technology:

Ollama & LLaVA: I used Ollama to run the open-weight LLaVA vision model entirely on my local machine. No API keys, no internet required.
Python: I wrote a simple CLI script (app.py) that takes an image path, converts it to base64, and sends it to the local Ollama client running on http://127.0.0.1:11434.
Backboard: I used my Backboard credits to review my Python code and optimize it for lower memory usage.
Render: I used my Render credits to deploy a simple static landing page for the project.
Tinker: In the future, I plan to use Tinker to fine-tune the model on regional bird species to make the AI even more accurate for my local area.

Why Does Open Innovation Matter?

Closed AI keeps us tied to the cloud. When you are out in nature without a signal, cloud-based AI is completely useless. Open-source AI, like the models I used here, gives power back to the developer. It allows us to build tools that are private, fast, and work anywhere on Earth. I can customize the model and run it offline. This is the power of open innovation.

The "Aha!" Moment

The best part was testing it. I turned off my Wi-Fi, loaded a picture of a local bird, and ran my script. The AI successfully identified the bird and gave me a safety warning about observing wildlife from a distance. It was incredibly rewarding to see open-source AI working flawlessly offline.

Prize Categories

Best Use of Render
Best Use of Backboard
Best Use of Tinker

Demo & Proof

You can view the live project landing page (hosted on Render) here:
https://touch-grass-ai.onrender.com

My Hacktoberfest Journey: Building an AI that touches grass so I don't have to

Sakshi Yadav — Sun, 11 Oct 2026 14:55:54 +0000

Hacktoberfest is all about open-source collaboration. When I saw the "Touch Grass" challenge prompt, I knew I had to build something that gets people away from their screens.

I present to you: The Offline Nature Identifier.

The Idea-
I wanted a tool that hikers, birdwatchers, and gardeners could use without a cell signal. A tool that runs entirely on your laptop, using open-source AI.

How I Built It (And The Tools I Used)

Ollama + LLaVA: I used these to handle the local vision AI. It identifies plants and animals right on your machine.

Backboard: I used my Backboard credits to review my Python code and make it more efficient.

Render: I used Render to deploy a simple static site for the project, making it accessible to everyone.

Tinker: In the future, I plan to use Tinker to fine-tune the model on regional bird species.

The "Aha!" Moment
The best part was testing it. I turned off my Wi-Fi, loaded a picture of a local bird, and ran my script. The AI successfully identified the bird and gave me a safety warning about observing wildlife from a distance. It was incredibly rewarding to see open-source AI working flawlessly offline.

Why Open-Source AI is Essential
Closed AI keeps us tied to the cloud. Open-source AI, like the models I used here, gives power back to the developer. It allows us to build tools that are private, fast, and work anywhere on Earth.

See the Project:
I deployed the landing page on Render:
[https://touch-grass-ai.onrender.com]

intro

Sakshi Yadav — Sun, 11 Oct 2026 06:33:27 +0000

Hi, I’m Sakshi Yadav. I’m a 3rd-year B.Tech Computer Science student at Gyan Ganga Institute of Technology and Science, Jabalpur. I’m interested in software development and currently improving my skills in C++, JavaScript, and DSA. I’m a quick learner, enjoy connecting with people, and I’m always looking for opportunities to learn, build projects, and grow as a software engineer.

intro

Sakshi Yadav — Fri, 09 Oct 2026 19:40:32 +0000