Success Story Garje Marathi AI mentoring

Garje Marathi AI mentoring: A pioneering initiative from the Garje Marathi innovation academy


Today, Artificial Intelligence (AI) is impacting all areas of our life.

I teach AI at the University of Oxford. I have been mentoring students and start-ups at Oxford for AI.

So, when Anand proposed to me the idea of creating an AI mentoring programwithin the Garje Marathi Innovation Academy, I was happy to set it up from a technical perspective.

We started with some simple guidelines for the AI mentoring program

  • A social / community led initiative
  • The mentorship program ran for three months
  • Work towards a demo in three months (specific objective for each participant)
  • Two weekly zoom calls
  • Work with real AI models(NLP, Computer vision etc.)
  • Create tools, best practice, templates etc. for future use
  • The start-up owns the intellectual property

Observations and Insights from mentoring

• We received a wide range of applications from farming, personal feedback,

fraud/anomaly detection

• Youngest participant only 19!

• All five participants completed the program and have a demo

• Selection criteria was based on practical feasibility(building a baselineproduct in three months)

• The initiative required time and effort with regular follow up


The selected initiatives and members were:

Farm Assistance Device for fruit farmers to improve farmproductivity while reducing use ofirrigation & fertilizers. – by Sanket Kedar. Objective: Develop AI & IoT based intelligent farm assistancedevice for farmers. Uses time series data

Object detection utilizing AI and Secure Edge ComputingByShreyas Panditrao. Objective: Use computer vision to detect the presence of hardhats (PPE) in the field in hostile environments(ex: mines). Uses computer vision on edge devices

Public Grant Prediction and Fraud DetectionBy Samir Gujar & Yuwaraj Shinde. Objective of this project is to learn the grant makingpattern/model for EU grants and Applications. Using thesepatterns/models we will be helping the Organization to catch thefraudulent attempts and to predict which application will begetting the grants. Uses natural language processing.

YouBetterMe Personal feedback tool –by Amol Parab and Akshay Sankpal. Objective - To enable important conversations and feedback. Uses natural language processing. 

To develop an anomaly detection toolBySagar Warke. Objective: Develop models for anomaly / outlier detection for IoT Data. Uses supervised and unsupervised anomaly detection models

(representative image of graph – source

Conclusions. Lessons learnt and Next steps

  • The program was a success and could serve as a model for future mentoring (AI or non-AI programs)
  • We need more women applicants for next time. That’s an area where improvement is needed
  • The next program will commence in January 2021
  • We will be investing in our own tools and platform to scale up
  • Many thanks to Anand Ganu and Ravindra Gadgil for the guidance!


Ajit Jaokar