BRAINSTORM. BUILD. BELIEVE. 

Github Link

In this project, we focused on a real world FMCG supply chain environment simulation created using OpenAI Gym and used Q-learning to train an agent that learns optimal reorder policies over time. We also implemented a Double Deep Q-Network and a random reorder point baseline policy. 

What fascinated me most was seeing how even small changes in the inventory environment, like lead times, demand variance, or warehouse constraints, could lead to drastically different outcomes in the agent's learning and performance.

This experience sharpened my understanding of the FMCG supply chain landscape, the nuances of RL algorithms in applied settings and a lesser explored but highly impactful domain: Inventory control using RL. 

Github Link

I developed a SQL-based inventory analytics system for a 2-year dataset across 5 stores and 5 product categories, uncovering insights on fast/slow-moving goods, seasonal demand shifts, stockout drivers, and supplier reliability. I implemented inventory turnover, reorder points, and safety stock logic to improve replenishment and identified clothing as the most profitable yet stockout-prone category. To make insights actionable, I built an interactive Replit dashboard with KPIs and modules for forecasting, regional trends, and store-level analysis, integrating data analytics, business insights, and visualization to guide better supply chain decisions. 

Github Link         Product Deck Link

I worked on EcoShop, a growth-stage e-commerce startup focused on sustainable and eco-friendly products, to design solutions that improve Gen-Z engagement and repeat purchases. I conducted user research, competitor benchmarking, and persona development, identifying key needs such as affordability, authenticity, and mobile-first experiences. Based on insights, I proposed a website revamp with features like trending product highlights, robust filters, reviews tab, subscription services, and real-time delivery tracking. I also suggested a tech stack (FastAPI, React, PostgreSQL) to support scalable filtering, and defined KPIs including bounce rate, repeat purchases, average session duration, and CTR. Finally, I built a clickable prototype and outlined a phased go-to-market plan to enhance customer retention, trust, and sustainable growth. 

Assignment Deck Link

I developed a strategic growth plan for Zusconnect, a travel creator marketplace, outlining initiatives to scale from the ground up. My work included creator segmentation, onboarding strategies, city scoring frameworks, and TAM–SAM–SOM market sizing, along with monetization models and phased GTM execution. I proposed multi-channel acquisition strategies (social outreach, campus programs, in-person events), designed creator scoring criteria, and identified key KPIs such as repeat bookings, CAC vs. earnings, and user retention. Additionally, I introduced an AI-powered “Agentic Search” concept to enhance discovery and booking conversion, with projections of up to 3x retention gains. To support execution, I mapped team requirements, risks, and escalation workflows, integrating analytics-driven insights and automation to strengthen creator success and platform scalability. 

Product Deck Link

I conducted a product teardown of Nykaa as part of Product Space Season 12, focusing on strategies to drive higher repeat purchases and customer loyalty. Through competitor benchmarking, user persona research, and pain-point analysis, I identified key issues such as cluttered landing pages, delivery delays, and lack of personalization. I proposed innovative solutions including hyperlocal logistics, agentic AI-powered search, AR-based virtual try-on, and face analysis for personalized recommendations, each evaluated with RICE scoring and MVP implementation timelines. This teardown was rated 8.2/10 by Product Space, validating the impact and feasibility of my proposed solutions to enhance user engagement, retention, and overall shopping experience. 

Product Deck Link

I developed Hermes, an AI-powered personal finance assistant targeting Gen Z and young professionals. Conducted a survey of 30+ respondents, identifying gaps in expense tracking (70% users) and investment awareness (60%). Prioritized features like UPI integration, Excel linkage, and AI chatbot using the RICE framework (highest score: 720K). Sized the $212M budgeting app market (2023) with 75% CAGR (2024–2033) and built a GTM strategy leveraging UPI partnerships and influencer trust. Defined North Star Metric as 4★–5★ user ratings tied to consistent savings. This project strengthened my skills in market research, feature prioritization, and go-to-market strategy. This deck qualified for the finals, ranking in the top 10 among 600+ participants nationally. 

Product Deck Link

I designed an Event Management Tool for IIT Jodhpur to streamline fest and event organization. Conducted user research via Google Forms and developed personas for festival chiefs (ages 19–23) and first-year student volunteers (ages 16–19). Identified key pain points like senate instability, late approvals, poor communication, and budget constraints. Created a user flow diagram for event proposals and scheduling, ensuring smoother collaboration between students and admin officials. Built wireframes for both student and admin interfaces, focusing on structured planning, transparency, and reducing delays. The solution improves efficiency, fosters collaboration, and enhances the overall fest experience. 

Github Link

Developing an agentic AI system that automates the design, training, and evaluation of deep learning models for turbulence super-resolution using the BLASTNet Momentum128 dataset. The agent reads dataset metadata, proposes candidate architectures (3D UNet, ConvLSTM, physics-informed variants), and iteratively improves model performance through automated experimentation. Performance is assessed via statistical error metrics and physics-based evaluations (energy spectra, divergence). This approach accelerates R&D workflows in turbulence modeling, reducing manual trial-and-error and enabling faster insights for aerospace, energy, and chemical process industries. 

Github Link

I am currently building an open-source Streamlit dashboard for HealthKart to track the ROI of influencer campaigns across Instagram, YouTube, and Twitter. The tool simulates datasets for influencers, posts, tracking, and payouts, and provides ROI & incremental ROAS analysis, performance insights, top influencer identification, and payout tracking. It also features filters by brand, product, persona, and platform, with options for data upload and export. This project highlights my skills in Streamlit, data modeling, product thinking, and insights storytelling.

Deck Link

I worked on a case study of RupayPay, a fintech lending startup, to evaluate whether it should remain bootstrapped or pursue a VC-backed path. I conducted market sizing (TAM ₹1,25,000 Cr, SAM ₹18,750 Cr, SOM ₹375 Cr) and a competitor analysis of BharatPe, Paytm, and Razorpay, alongside a SWOT assessment highlighting strengths like 70% CAGR and low NPAs, and threats such as regulatory pressure and competition. Through scenario analysis, I modeled bootstrapped growth at 50% CAGR with stable profitability versus a VC-backed path at 150% CAGR with higher burn but rapid scale. I also assessed risks across operational, financial, cultural, and regulatory dimensions, while proposing AI-driven personalization, credit controls, and proactive regulatory strategies. Based on the analysis, I recommended the VC-backed path for achieving scale and long-term competitive advantage.

Deck Link

I worked on a case study challenge by Swiggy, a leading food delivery app, to look for growth and expansion ideas on my college campus. I came up with 8 new innovative ideas for Swiggy to expand in college campuses. I conducted surveys across my campus through a Google form and in person interviews with various stakeholders. I elaborated in detail about one of my ideas: Swiggy Sync, which enables students to share cart in real time and notifies everyone of when an order is being placed on campus. I proposed a detailed GTM plan with a cost benefit analysis which concluded that Swiggy Sync would reach breakeven in 9 days within launch of the Phase 2 plan.  

As part of my Introduction to Machine Learning Course Project in Semester 3, we worked upon a limited Kaggle dataset on AI-powered Job Market Insights. I intend to revamp the project by creating a Linkedin Jobs dataset, analysing the job market, identifying trends by 2030 and design a Career Tech product for recruiters focused on skill based hiring. 

My favorite product is Spotify which has tried and tested many features in the past. Here, I will analyse Spotify and its global presence. I will vibe code a working prototype of my feature which involves creating a Karaoke option for users (individual and group) and estimate its product-market fit. I will also analyse the current AI voice recognition features and legal compliance factors associated with its implementation.