Achievements

March 2021:

In this globally contested Data Analytics competition hosted by the prestigious Wharton School, our team consisting of Nischith Sriram (team leader), Luv Nambiar & myself secured first position. As the only Indian and undergraduate team among the finalists, we also got an opportunity to present our insights at the Wharton People Analytics Conference 2021.

The objective of the competition was to provide actionable recommendations to the 'Teach For America' nonprofit organization. Our eventual findings involved using extensive & creative data analysis & nudge theory to make data-driven suggestions to TFA's applicants. We also proposed an ambitious and innovative Reinforcement Learning model to make optimized offers to the TFA applicants. My own contributions were towards data analytics & visualizations.

For more details about the competition & our solution, you can watch our Global Finals presentation. Feel free to watch others' presentations as well -

This is our final presentation which was shown at the Wharton People Analytics Conference 2021 -

February 2021:

In the first edition of this international hackathon hosted by India and the ASEAN countries, our multi-national team consisting of Keerthi Rajan (team leader), Shruti Rampure, Swapnil Panwala, Ngoc Bui, SivEng Orn & myself and mentored by Randall Sie & Mya Mya Mhway, secured first position for our problem statement.

Our problem statement concerned defence against pirate attacks on the sea. Our solution was threefold: Risk Analysis before ships set sail based on historical pirate attack occurences, Piracy Defence while ships are sailing using several hardware sensors and Post-Attack Tracking & Insurance Evidence based on satellite imagery. My own contributions were towards Risk Analysis & Satellite Image Processing.

For more details about our solution, you can view our Finals presentation describing our proposed safety features -

August 2020:

Our team consisting of Saurabh Agarwala (team leader), Vaibhav G., Shashank Jaiswal, Linu george, Rakshatha Vasudev & myself and mentored by Salman Shah & Dr. Sowmya Kamath S. secured first position for our problem statement in the nationally-contested Smart India Hackathon hosted by the Ministry of Education Innovation Cell.

Our problem statement was regarding cotton price forecasting. Not only did we perform extensive feature engineering and build different Time Series Forecasting models capable of forecasting cotton prices for a given district & state on a daily, weekly and monthly basis, but we also developed an e-commerce website and android app to complement the application of wisely using cotton forecasting to buy or sell cotton produce.

For more details about our solution, you can view our Finals presentation describing our features, motivation and tech stack -

May 2023:

Organized by Rolls-Royce's R2 Data Labs and hosted on HackerEarth, I was selected as the winner of this national-level hackathon.

The hackathon had an open problem statement about helping airlines identify new air travel routes in the Indian market. My proposal was called 'AIRDASHER' standing for "Analytical Intelligence for Route Development and Strategic Handling of Economic Resources" - as the name suggests, it served as an end-to-end product for airlines to identify new air routes, integrate it into their existing flight network and perform comprehensive cost & resource analysis about the viability & profitability of such new routes.

I had developed a web interface where tier-I & tier-II cities were modeled based on economic, social, educational & tourism factors to pinpoint exciting new destinations for airlines. Once integrated into the airline's network, depending on different parameters like demand, flight costs, operational expenses, etc., airlines could analyze and tweak parameters to justify viability and determine profitability of starting such new flight routes.

For more details about our proposal idea, you can watch our idea description video -

February 2022:

A national competition organized by NEC, Mitsubishi Corporation India & Isuzu Motors India and hosted on HackerEarth, I was selected as the winner of the hackathon after multiple rounds of preparation, presentation & shortlisting.

The hackathon had an open problem statement about unique & innovative ways to utilize, enhance or market Isuzu's pickup trucks for the Indian market. My proposal was called 'The Rurban Traveller' - it aimed at marketing Isuzu's passenger pickups as the ideal 'Rurban' (rural + urban) vehicle since it was powerful & durable for rural terrains and also stylish & useful for urban environments.

Complimenting this marketing would be a mobile app - connected through IoT to an OBD2-based embedded device in the pickup. Routine mode of this mobile app included features like smart vehicle-usage monitoring and fuel-level alerts. Adventure mode included features like adventure recommendations & automatic compilation of a trip's images & videos in a visually wonderful graphic called 'Memories'.

For more details about our proposal idea, you can watch our idea description video -

February 2021:

For this nationally-contested hardware challenge hosted by the Ministry of Electronics & Information Technology, Government of India, our team consisting of Atulya Mahesh, Kumar Alabhya, Sahith S R, Abhijit Gupta & myself (team leader) and mentored by Dr. Ramesh Kini reached the Semi-Final stage. We were among the 100 teams throughout the nation to do so. Aimed at promoting startups, the finalists of this challenge will get the chance to be a government-funded startup.

The objective of the challenge was to propose and build an application using the Industry-grade Swadeshi Microprocessor families called Shakti & Vega. Our proposal was called 'Smart Road Traffic Control & Monitoring System'. We proposed using CCTV cameras to dynamically schedule traffic signals based on the real-time traffic at an intersection. Instead of following the pre-determined cyclic sequence of traffic lights, the traffic signals would be able to detect the level of traffic on each road and then decide which traffic lights to change and for how long.

Other innovative features include identifying, locating & emphasizing passage of Emergency Vehicles, Inter-city Communication to control city-wide traffic, Noise-based Penalties to discourage noise pollution, Traffic Violation Detection along with number plate recognition, Accident Detection with immediate notification to relevant authorities, Smart Pedestrian Crossings for the traffic signal to also consider pedestrians waiting to cross a road and Real-time Ground-level Traffic Updates which would be much more reliable than satellite-based traffic updates.

For more details about our proposal idea, you can watch our idea description video -

March 2021:

A national competition organized by NEC & Mitsubishi Corporation India and hosted on HackerEarth, I was among the top 20 finalists qualified for the final hackathon.

The hackathon had an open problem statement about improving the service, marketing & retention rate of highway-side amenities called 'CubeStops' through technological solutions. My proposal idea was called 'AI-boosted Online Ordering for FoodStops'. My proposal would enable highway-travelers to pre-order their food before physically reaching the CubeStop.

Using a virtual menu, customers could order and have their food ready before they reach the CubeStop to minimize the time of their journey. The menu also makes recommendations to customers based on their order history. To make recommendations, I designed 'ingredient embeddings' which can represent different food items based on taste, origin, type of dish (entree, main course, dessert), etc. I also developed an approach for customers to provide extremely quick feedback by simply clicking a picture of their food and giving a star rating. Based on the picture, my models can identify the food item and accordingly manage the review.

July 2021:

A national competition organized by NEC & Mitsubishi Corporation India and hosted on HackerEarth, I was among the top 20 finalists qualified for the final hackathon.

The hackathon had an open problem statement about adding value to Indian middle-class residential properties through technological solutions. My proposal idea was called 'Reassurance Chatbot'. My proposal involved developing a personalized chatbot to provide reassurance to customers who have invested in under-construction projects.

The chatbot can not only answer simple general questions of the customers, but is also capable of providing a personlized experience through the use of 'NL-to-SQL'. Essentially, the chatbot can answer customer-specific personal questions by converting the question to an SQL query and then fetching the answer from a database. The chatbot can also help provide customers with real-time updates about the construction to reassure them about the risks of under-construction projects. Using the current progress, it is also capable of forecasting the date of completion using multiple Time Series Forecasting models.

September 2023:

An international competition of WowDAO, I was among the semi-finalists for the Synthetic Data Applications theme.

The hackathon had an open problem statement about innovating sythetic data usage in sectors with limited real datasets. My proposal was called "Signalytics" which was a simulation & analytics tool for testing different kinds of dynamic traffic signal scheduling algorithms for various traffic conditions.

Since dynamic traffic light scheduling can drastically improve traffic at busy intersections, Signalytics provided an interface where algorithms can be quantitatively judged across several unique metrics. Also capable of visually simulating algorithms in different kinds of traffic conditions, Signalytics can help in bringing up confidence when introducing such algorithms to the real-world.

April 2021:

I secured a top 30 ranking in this national-level Machine Learning hackathon. The competition was organized by Univ.AI.

The problem statement involved predicting the credit risk score of different individuals. My approach involved extensive feature engineering, SMOTE resmapling to handle the class imbalance, thorough hyperparameter tuning and ensembling gradient boosting models like XGBoost, LightGBM & CatBoost.

February 2020:

In this global presentation competition organized by IET, I reached the Local Network Round held in Bengaluru and secured a top 4 position. Participants were needed to present & deliver a presentation based on any technological topic.

My topic of presentation was 'Future of Artificial Intelligence in Healthcare'. My presentation touched upon the fear that machines might replace doctors & nurses in the future. However, my argument was that simply obtaining better performance than doctors cannot be a justification to replace doctors; since emotions and feelings like empathy, human touch & intuition are essential aspects of a doctor's job.

April 2021:

Due to the cancellation of IET PATW 2021 because of the Covid19 pandemic, the student branch of IET at NITK Surathkal held this online presentation competition. Participants would be judged online based on presentation, slide content and delivery.

My topic of presentation was 'Neuromorphic Computing'. My presentation explained the fundamentals of neural networks and software-based training of these networks; and then introduced the concept of Neuromorphic Computing where hardware tries to mimic neurons, thereby eliminating several limitations of conventional software-based training.

October 2018:

I secured second postition in this local Competitive Machine Learning contest. The contest was held by IEEE NITK, the student branch of IEEE at NITK Surathkal. The problem statement involved predicting a customer's star rating based on the same customer's textual feedback.

My solution involved developing word embeddings through Word2Vec and applying them on an ensemble of gradient-boosting models like XGBoost & LightGBM.