Projects

Challenge

India's 12M+ gig workers, with 60% being new-to-credit and lacking formal income proof, faced a ₹66,000 Cr credit gap, forcing reliance on predatory informal loans and leaving a massive market segment underserved by traditional lenders.

Action

Solution: GigScore™ AI Engine, an alternative data-driven underwriting system, to provide instant disbursal and flexible EMI options through app-first onboarding, by integrating UPI flows, platform ratings, GPS patterns, and enriched bureau data, all within an RBI-compliant Account Aggregator Framework.

Result

4-hour loan disbursal at 18-24% APR, thus opening formal credit access for millions and creating a ₹1.8L Cr Total Addressable Market.

Challenge

Limited access to AI services on LinkedIn for those who need it most, such as freshers, due to the cost of the Premium plan

Action

Built a personal portfolio website with AI integration across most aspects of the site, including a careers section with AI-driven ATS screening, resume building, and job search

Result

Created a personalized alternative to LinkedIn Premium with AI-powered features

Challenge

Lack of tools to plan courses in the institute's ERP system, leading to confusion and primitive methods like pen and paper or Excel being used by students.

Action

Created a platform, https://www.academic-planner-iitgn.vercel.app, using AI-assisted development.

Result

A platform was made to fill the vacuum of course planning tools.

Challenge

Indian retail firms faced intense pressure from low margins and high volume, significant inventory risk, supply chain inefficiencies, and fragmented decision-making, making sustainable profitability a complex challenge. The core problem was understanding how a data warehousing-enabled Management Information System (MIS) could transform operational control into consistent, sustainable profit.

Action

Analyzed DMart's business model and MIS strategy, examining its data warehousing approach, inventory and supply chain decisions, and SKU rationalization process through a strategic management information system lens. Developed key learnings about the critical role of data discipline and operational control in retail profitability.

Result

The analysis identified that DMart achieved a 15% Return on Capital Employed (ROCE) advantage over traditional retailers, maintained over 60 days sales before vendor payments, and operated with 30-40% fewer products than competitors. Strategic MIS proved to be a profit governance tool, not merely a reporting system, driving sustainable margins and cost leadership.

Challenge

I come from the northeast (Silchar and Agartala), and we were introduced to the concept of malls rather recently. It was only in 2012 that the Goldighi Mall opened its doors in Silchar after a long construction time, followed by the ML Plaza in Agartala in 2013. Interestingly, both had BigBazar as their opening partners and for a long time, the mall was colloquially known as BigBazar. These malls instantly became the heart of their respective towns, attracting massive footfall in the initial days and, to some extent, becoming the envy of local store owners. It is worth noting that these were malls in the true sense, built over a large area, hosting multiple brand outlets (albeit not so high-end), a gaming plaza, a food court, and a cinema complex. The sunshine days for these establishments started to wane about half a decade into their operations. Things were already slowing down, and then the pandemic became the last nail in the coffin. Although neither of them has stopped operations today, their footfall is a shadow of what it was. What were once bustling centres of commerce have almost been abandoned. I observed a similar cycle in Bengaluru and Ahmedabad, and they were much more pronounced. Unlike towns in the northeast, Bengaluru and Ahmedabad are metropolises with malls sprouting on every other corner and every other day. But these new malls stand in contrast to older malls that have shut their doors. Seeing this, a question came to my mind: Why do malls die? What are the most important factors? What is the optimal density of malls for a given population? Establishing and running a mall requires significant investment, and many families' livelihoods depend on it. Hence, this question becomes significant. I have started my investigation, and I hope to have an answer soon. If anyone is interested in contributing to this work, you're welcome to reach out to me.

Challenge

A core challenge involved accurately modeling the 2D incompressible Navier-Stokes equations for lid-driven cavity flow, specifically mitigating pressure-velocity decoupling inherent in co-located grid discretizations to achieve stable and realistic fluid dynamics.

Action

Engineered a Finite Volume Method (FVM) solver in MATLAB, discretizing the incompressible Navier-Stokes equations using explicit Euler for time marching and applying Upwind Differencing (UDS) for convection terms and Central Differencing (CDS) for diffusion. Implemented Rhie-Chow interpolation to prevent pressure-velocity decoupling, constructing a sparse Laplacian matrix for the Pressure Poisson Equation (PPE), which was solved iteratively until a 10^-7 velocity tolerance was achieved.

Result

The simulation converged to a steady state at Reynolds number 100 on a 101x101 grid, yielding velocity and pressure profiles consistent with standard benchmarks and accurately visualizing the dominant clockwise vortex and corner eddies.

Challenge

Ambiguous problems surrounded the effectiveness of social media influencer campaigns, leaving many companies unable to reliably measure return on investment despite increasing marketing budgets.

Action

Designed and executed a mixed-methods marketing research study, developing a framework that integrated exploratory qualitative insights with structured quantitative surveys from a 100-120 respondent sample to define and predict influencer effectiveness.

Result

The developed framework provides a structured approach to analyze the psychological drivers of influencer marketing, isolating causal insights on brand perception and consumer behavior, moving from an ambiguous problem to clear, measurable outcomes.

Challenge

A fundamental engineering problem involves predicting the intricate dynamics of a serially connected N-mass, N-1 spring system, constrained between fixed boundaries and subjected to external forces, requiring the formulation and computational solution of N coupled, second-order ordinary differential equations.

Action

Engineered a comprehensive numerical simulation in MATLAB, deriving governing equations from Newton's Laws of Motion and formulating them into a state-space representation, then implemented a fourth-order Runge-Kutta (RK4) solver for stable and accurate time-domain integration.

Result

The developed model accurately captured complex dynamic phenomena, including coupled free vibrations, steady-state forced response, and longitudinal wave propagation, confirming the numerical solver's stability and correctness across various test scenarios.

Challenge

The primary challenge involved developing a program to accurately simulate unsteady, inviscid, compressible fluid flow through a converging-diverging rocket nozzle, and then computing steady-state flow properties across both isentropic subsonic and subsonic-supersonic flow regimes. This required correctly capturing potential discontinuities and conserving mass, momentum, and energy across the domain, all while maintaining numerical stability and accuracy.

Action

Engineered a program in MATLAB to simulate quasi-one-dimensional, unsteady, compressible fluid flow through a converging-diverging rocket nozzle by implementing the MacCormack scheme to solve Euler equations. Designed and applied appropriate boundary conditions and a time-marching loop for convergence across two distinct isentropic flow regimes.

Result

The numerical results for pressure, density, temperature, and Mach number profiles demonstrated strong agreement with analytical solutions and a very close tie to NASA experimental data, validating the model's ability to accurately capture real-world fluid behavior across both isentropic subsonic and subsonic-supersonic flows.

Challenge

A dynamic and highly competitive Indian chocolate and confectionery market, characterized by diverse consumer preferences, significant competition, and varying price sensitivities across urban and rural demographics, posed a complex challenge for Nestlé India to maintain and grow its market share.

Action

Spearheaded a secondary research-based critical analysis of Nestlé India's Chocolates and Confectionery division, establishing it as a Strategic Business Unit (SBU). Evaluated its marketing mix (Product, Price, Place, Promotion), market environment (5Cs), and consumer segmentation, targeting, and positioning (STP) to identify key performance drivers and strategic opportunities.

Result

The analysis provided a comprehensive understanding of Nestlé India's market position, identifying significant growth drivers such as new product variants contributing 6.5% to sales and deep rural penetration reaching over 108,000 villages. Strategic recommendations were formulated to enhance health-oriented and premium offerings, customize for rural tastes, and amplify digital marketing efforts.

Challenge

A critical challenge involved accurately modeling the two-dimensional unsteady heat conduction equation on an 81x81 square domain with mixed boundary conditions, requiring the development of robust numerical schemes to solve for the temperature field's time evolution until a stable steady state was achieved.

Action

Engineered computational solutions by discretizing the governing parabolic PDE using the Finite Difference Method, implementing and comparing explicit, fully implicit, and Crank-Nicolson time-stepping schemes, and integrating a Tridiagonal Matrix Algorithm (TDMA) for implicit solutions to solve the system of algebraic equations.

Result

Achieved convergence to the same steady-state temperature distribution across all three schemes; the Crank-Nicolson scheme notably reached steady state in 7,890 steps (789.0 s), outperforming the explicit (20,163 steps, 907.3 s) and fully implicit (9,529 steps, 952.9 s) methods in terms of computational steps and time.

Challenge

Steel components in harsh-service environments face premature wear, corrosion, and replacement costs. Tata Steel’s objective was to evaluate whether WAAM-based SS316L deposition over mild steel could create a defect-free functionally graded repair solution that improves component life while reducing refurbishment cost.

Action

Conducted a systematic literature review of 250+ sources to identify research gaps in WAAM-based FGMs, then designed and fabricated MS–SS316L graded structures (200 × 30 × 110 mm) using a 6-axis CMT WAAM robotic system. Performed metallurgical and mechanical characterisation using optical microscopy, SEM, hardness contour mapping, and tensile evaluation, while documenting findings in a 20,000-word technical completion report delivered on time.

Result

Developed a defect-free hybrid steel repair strategy with strong metallurgical bonding and gradual hardness transition, achieving 40% higher hardness, 20% improvement in tensile strength, and 10% projected cost reduction, demonstrating direct applicability for marine, oil & gas, and heavy industrial asset life-extension programs.

Challenge

The practice of burning rice husks caused significant environmental pollution and waste, while 3D concrete printing faced challenges in developing sustainable, yet high-performance, printable concrete mixes that meet both fresh and hardened property requirements.

Action

Led a team of 4 to develop a factorial experimental design to systematically vary proportions of rice husk ash (RHA) and polyvinyl alcohol (PVA) fibres within concrete ink formulations. Conducted comprehensive tests, including slump, flowability, buildability, compressive, and split tensile strength analyses on 3D printed specimens to evaluate fresh and hardened properties.

Result

The optimised Mix M3 (10% RHA and 0.5% PVA) demonstrated the best balance of printability and mechanical performance for 3D concrete printing, achieving 31% lower slump for improved shape retention, while simultaneously delivering 4% higher flowability to preserve extrudability. The mix also enabled a 20% reduction in cement usage, while exhibiting 15% higher strength than the base mix. Communicated to JOM.

Challenge

The challenge was to understand and predict how microstructural randomness, such as variations in platelet sizes and interface properties, impacts the mechanical behavior and failure mechanisms of nacre, a complex biological composite material.

Action

Engineered a comprehensive finite element model (FEM) in ABAQUS 6.14, integrating Tension-Shear-Chain (TSC) and Virtual Internal Bond (VIB) concepts, to simulate nacre's deformation and failure under uniaxial tensile strain with periodic boundary conditions. The microstructural randomness was quantified using the Weibull distribution function.

Result

Demonstrated that nacre's brick-mortar microstructure effectively suppresses the adverse effects of microstructural randomness, rendering the ensemble strength less dependent on representative volume element (RVE) size and significantly enhancing mechanical robustness. The symmetric overlap ratio (11/2l = 0.5) was identified as most effective in mitigating randomness effects.