Hi, I am Aditya Singh.

Resume

I am an independent research professional and strategy consultant. Prior to this, I worked 3 years at KPMG's Research division within the Global COO Office team and worked on consulting and value creation projects and contributed to research papers, industry packs, company analysis, benchmarking, along with research-oriented content creation (press articles, newsletters, etc.).
I completed my Bachelor's in Commerce (Honours) with a Minor in Economics at University of Delhi (2019-22), where I mostly studied business, economics, and finance. During this period, I also cleard CFA Level 1, and interned with prominent companies, namingly, Deloitte, Mazars, and Vahdam India.
Since school days, I wanted to get into research and pursue PhD eventually. I enjoyed the research work I did during my internship at Mazars and was fortunate to find an opportunity with KPMG where I developed an interest in researching on MSME financing and have taken up some independent research in spare time. Beyond research, I love playing and watching Football (Forca Barca!) and enjoy playing guitar as a mean to destress.

To learn about my career journey, click the PDF link below (updated September 2025):

Open my CV (PDF)

Working paper: Did Liquidity Windows Fuel MSME Credit?

Research@KPMG

A sampler platter of research work I can publicly share (sanitised, of course). Working at KPMG was bittersweet because I worked with great clients on important pieces and learnt a lot from subject-matter experts of their fields, however when it came to contributions I always got crushed under hierarchy and "brand > employee" policy. Alas! Here are a few key projects:

• Tri-Criterion Evaluation of Generative-AI Use Cases in Financial Services: Integrating Impact, Feasibility, and Responsible-AI Risk Across Banking, Insurance, and Asset Management – We construct sector-specific value chains for banking, insurance, and asset management and identify 51 distinct Gen-AI use cases. Each use case is mapped to one (or more) of five canonical Gen-AI capabilities (i) information summarization, processing, and synthesization (ii) process automation, (iii) employee assistance and productivity enhancement, (iv) responding to and interacting with humans, and (v) data analysis and insight generation.
  Impact is operationalised via a bottom-up task analysis of 468 occupation-level activities drawn from U.S. Bureau of Labor Statistics and O*NET datasets. Tasks are tagged with a binary relevance flag for every capability; cumulative flags yield a 0–5 impact score subsequently normalised into low, medium, and high categories. Feasibility is captured through an eleven-item cost-and-complexity checklist covering data-quality remediation, model fine-tuning, infrastructure scaling, regulatory documentation, and downstream integration. Fewer items flagged correspond to higher feasibility; scores are again rescaled to a 0–5 continuum.Responsible-AI risk is quantified using an adapted Monetary Authority of Singapore (MAS) FEAT* checklist that evaluates fairness, explainability, accountability, and transparency hazards such as hallucination, data drift, and model inversion. Higher counts of potential violations produce darker chromatic intensities in the study’s visualisations.The resulting tri-dimensional dataset is rendered in four scatterplots: an aggregate view of the five overarching capabilities and sector-specific plots depicting individual use-case positions. Findings reveal a pronounced cluster of high-impact, high-feasibility, low-risk opportunities in document summarisation and employee augmentation, while predictive-analytics use cases exhibit elevated risk despite significant impact potential.
  The framework offers a reproducible decision tool for chief data and risk officers, enabling capital allocation that balances economic return with stakeholder trust. By unifying occupational task analytics, cost engineering, and responsible-AI governance, the paper advances multi-criteria evaluation methods for emerging technologies in regulated industries.
• Forecasting Indonesia’s International Bandwidth Gap: A Structural Demand–Supply Analysis of Submarine Cable Capacity – Indonesia’s ascent as a digitally enabled economy hinges on the timely expansion of its international bandwidth infrastructure. Using a structural demand–supply framework, this paper estimates Indonesia’s current and prospective gigabit-per-second (Gbps) requirements and contrasts them with the nation’s installed and planned submarine cable capacity. First, we construct a granular supply inventory of all active and announced submarine systems directly landing in, or transiting through, Indonesia (n = 65), drawing capacity figures from publicly reported design constraints, Telegeography route data, and operator disclosures. To isolate domestically available bandwidth, we deduct transit allocations earmarked for third-party markets. Second, bandwidth demand is forecast through 2032 via a bottom-up model that links per-capita international traffic to GDP per capita, smartphone penetration, data-intensive application adoption, and enterprise cloud-migration rates. The findings imply that (i) incremental capacity upgrades are insufficient; (ii) at least one additional high-fiber-pair cable with ≥ 15 Tbps lit capacity is economically justified within the next five years; and (iii) policy incentives that expedite permitting and spectrum landing rights could mitigate impending congestion. Beyond corporate strategy, the paper contributes to the literature on digital infrastructure bottlenecks in middle-income economies and offers a transferable methodological blueprint for emerging markets.
• Quantifying the Modular-Building Fleet in Australasia: A State-Level Supply Gap Analysis for Social Housing Policy – Australia and New Zealand are turning to volumetric and panelised modular construction to accelerate the delivery of affordable and social housing. Yet policymakers and private developers lack an inventory of deployable modular assets the “fleet” that can be mobilised at short notice. This study develops the first systematic census of relocatable modular-building stock across all eight Australasian jurisdictions (the six Australian states, the Northern Territory, and New Zealand).
  We begin by identifying 27 active market participants through company filings, rental registries, tender documents, and satellite-imagery verification. For each firm we disaggregate owned or contracted units into three functional archetypes: (i) build-own-operate (BOO) villages, (ii) temporary site accommodation, and (iii) permanent or semi-permanent prefab homes. Fleet counts are normalised to “equivalent single-module units” to enable cross-type comparison.
  The findings indicate that: (1) existing fleets are heavily concentrated in resource-rich site-accommodation assets unsuited to urban social-housing needs; (2) inter-state relocation alone cannot close the gap without significant retrofit investments; and (3) accelerated procurement of purpose-built modular homes potentially via public-private partnerships offers the fastest route to bridging deficits. By providing a replicable fleet-quantification framework, the paper informs infrastructure planners and housing authorities seeking data-driven allocation of modular assets in the face of acute affordability pressures.
• Several complex company level analysis - The role at KPMG also included deep diving into singular market participant, often times operating within new or unheard sectors which nonetheless required me to learn about them not only comprehensively but also quickly. For instance, one of KPMG's client worked in Legacy Acquiring sector which belongs to niche sub-sector of insurance. It quintessentially operates like a private equity firm that goes for distress purchase or carve outs. Insurance companies sell their legacy portfolioes lying on their balance sheet as bad premiums to companies like my client. Our main advisory service extended was a loss ratio (claims and expense) analysis, which included ways to transform sub-optimal portfolios and manage other operating expenses. But how do you do that when the entire business model resided on buying losses!

Personal Projects

I like building things that scratch my own itch or teach me something new. A few highlights:

• Algo trading app (work in progress)  – I put my Python skills to test and tried to build an algo trading bot that predicts the stock price of PVR Inox and makes automated trade executions. Used YFinance, Statsmodel, Pandas, Matlplotlib libraries mainly to leverage statistical tools such as ARIMA, GARCH, and LSTM. It worked great for some time and then I overkilled it and have to debug the messy code. Its a work in progress.
• Mayfair Rugs  – In August 2023, I founded a UK-based premium hand knotted rugs brand, "Mayfair Rugs", that linked rural weavers in India from ~10 villages to high-end point of sales from designers to international brand. I started this because I "knew" social enterprises as I setup one during my undergraduation. However, this one was way tougher challenge since supply chains were crippling with middlemen. I took upon me to solve this by disintermediation and finding moderators from the weaving communities to become solo-preneurs within Mayfair ecosystem. They were accountable for production, inventory (final, semi-finished, raw material) and communicated with production team in Delhi. With this we uplifted the livelihoods of the weavers and the community while improving the income statements of overall Mayfair Rugs. We sold to retailers in the US and collaborated with UK based designers and international home-decor brands. Some deals went through, some didn't. I even hired a Marketing Manager in London who helped us build connections with buying houses and brands in the UK and represented Mayfair in global trade expos. Overall I learnt a lot about building businesses, unit-level costing and economies, marketing, hiring, and making key decisions.
• Webscraper for retrieving latest articles  – Using Selenium and Webdriver I created a headless bot that ran Google searches on Chrome and retrieved links of a pre-specified period published by an input company/organization/any entity and provided final output in an Excel using Pandas. The tool is a great value addition for researchers within competitive intelligence teams to identify white spaces of research or for data collection by academic researchers such as "retrieve links for Financial Stability Report published by South Africa Reserve Bank from 2020-2025".
• Financial Model of NU Holding  – Built a dividend discount model (DDM) and discounted cash flow (DCF) model of Nu Holding Bank on MS Excel. I made this after listening to David Velez invterview conducted by Stanford Business School. I created a fully dynamic 3-statement financial statement of NU Holding on MS Excel and took key financial assumptions to forecast the future free cash flows of the company. And finally discounting them using the calculated WACC of the firm. For DDM I estimated the cost of equity using CAPM and Gordon Growth Model to estimate the share value.