31 docs tagged with "Data Science intern, Syngenta"

The collected images were analyzed by the computer vision model to detect crops, area coverage of crops was computed and the percentage coverage of each crop was returned.

The first step is to break down the given prompt into a sequence of simpler prompts that are easier to process.

Now that our forecasting is complete, we can compare the forecasts with IBP and get the difference in accuracy and errors.

The dashboard for presenting the results of this project was made programmatically using the Tremor framework of React Javascript. The web app was made using the NextJS framework and TypeScript.

The dashboard for presenting the results of this project was made programmatically using the Tremor framework of React Javascript. The web app was made using the NextJS framework and TypeScript.

Another page was added to the Sales Forecasting dashboard for Smart Supply.

Tech stack

Delivery data

The data required for this was the forecast data of both my forecasts and IBP forecasts from earlier, and the current inventory data which was fetched with the SQL query:

An llm does not have the innate capability to make dynamic visualizations like charts and KPI cards.

Choosing a good evaluation metric for our problem is important. How accurate our estimate is depends on what metric we are using, and depending on the situation, different metrics can be the most suitable.

After cleaning and preprocessing the data, we created the relevant and important features for forecasting.

The response from the tool used to process the final instruction may not be human-friendly, only containing a single number or such.

Now that we have the cleaned dataset with the appropriate features, we can finally perform forecasting. This forecasting is just regression, with the exception that train data must be sequential upto a date and test data must be the sequence following that.

This project is still a Work In Progress.

This project is a Work in Progress.

A custom GPT chatbot was made for each of the projects as an addon to the corresponding dashboard, including for Inventory Management, Document Retention and Crop Identification. From the master dashboard, the user can either go to the dashboard or the natural language interface.

This was my first internship project inspired from the pre-existing Field Segmentation project for a hackathon.

I am a Data Science intern in Production & Supply (P&S) department of Syngenta INPU (India Pune). The team, named Center of Expertise (COE), was responsible for the supply chain and logistics aspects of the AMEA (Asia Middle East Africa) region.

This project was about optimally managing the inventory of various products, particularly Crop Protection (Insecticides, Fungicides, Herbicides, etc).

This was a project started as an experiment to see whether using covariates like weather and economic indicators provide better forecasts than simply using historical data.

Now that demand has been forecasted and is better than IBP forecasts for a significant number of products in many countries/regions, I decided to make use of it to further recommend how inventory should be managed.

The data science team had made a number of end-to-end projects including the ones documented here. We needed a unified interface to interact with all of them since dashboards were cumbersome to both build and use.

The model used for SynBot was GPT35-pss. It was made using Azure OpenAI and LangChain.

There are several steps followed for preprocessing the data.

A prompt template is made with LangChain for prompt engineering.

The chat history must be passed to the cleaning agent for context. In case of a series of instructions rather than a single one, the responses to the intermediate tasks may be required as context for the subsequent tasks, so also must be passed as history.

The model for answering the prompts may require specific data to answer the queries. For example, if asked a question about forecasts, the model requires the actual forecast data.

For every plant, its forecast and IBP forecast are summed for the given number of months and compared to the total inventory.

The type of the prompt is detected next. This is an individual prompt in the sequence of instructions returned by the cleaning tool.

When the forecast accuracy got from our method is compared to IBP forecasts, we get some brands for which we have significantly higher accuracy/lower errors. We would like to improve the accuracy even further for these brands.