Strategy and technology

Thomas K Varghese

PGPM, SPJIMR ’27. Four years at EY R&D.

I break down ambiguous business problems and connect strategy with execution.

Explore my work View resume

Proof points

Results, not adjectives.

Collaboration capacity

Supported users rose from about 10 to 25.

3rd

of 45 teams

CONFLUENCE 2026.

EY Spot Awards

Performance and delivery.

2/3

AI PoCs advanced

To evaluation or development.

<1 day

K1 tax workflow

A multi-week cycle, shortened with human oversight.

Colleagues trained

Robotics and NVIDIA tools, Senior Associate to Assistant Director.

Co-authored research, Materials Today: Proceedings · Microsoft Azure Fundamentals · SHRM People Manager Qualification · Black belt, 7+ years in martial arts

Selected work

Each case starts with the outcome.

Consulting · CONFLUENCE 2026

Placed 3rd of 45 with a profit-led growth plan.

Our team identified an underserved HNI segment and built a differentiation and retargeting strategy.

Situation: A business needed a feasible path to improve profitability and reach an underserved segment.
What I did: Identified an underserved HNI segment and developed a differentiation and retargeting strategy with the team.
Result: Ranked 3rd of 45 teams. Projected ≈ ₹1.27 crore incremental profit (~58% uplift). Projections, not realised results.

Infrastructure efficiency · EY R&D

Found the bottleneck that capped a platform at 10 users.

A Unity WebGL collaboration experience became unstable past ten people. The fix took it to twenty-five.

Situation: The collaboration experience became unstable beyond approximately 10 concurrent users.
What I did: Identified a state-synchronisation bottleneck and reduced non-critical updates from roughly 60 Hz to 10 Hz, with client-side interpolation.
Result: Supported capacity increased from approximately 10 to 25 users (2.5×).

Process transformation · EY R&D

Two of three AI proofs of concept moved forward.

I led frontend delivery for AI proofs of concept, including document processing and a K1 tax workflow.

Situation: Enterprise teams needed usable interfaces to evaluate AI capabilities and streamline document-heavy workflows.
What I did: Led frontend delivery for three AI proofs of concept and demonstrated them to senior stakeholders.
Result: Two of three PoCs advanced to evaluation or development. The K1 workflow shortened a multi-week cycle to under a day, with human oversight.

Industrial transformation · EY R&D

Remote robot inspection for sites people shouldn't enter.

I built the remote-control and live-data tools for a quadruped inspection robot, letting stakeholders watch and act from anywhere.

Situation: Hazardous-site inspections needed faster preparation and live visibility for stakeholders who could not be on site.
What I did: Connected the robot to remote control over IoT, built a control website with live image streams and real-time AI inference, and solved a no-Wi-Fi networking constraint under a tight deadline. Contributed technical and business inputs to the internal funding case.
Result: Designed to cut inspection preparation from 15–20 minutes to under 5, from anywhere. Contributed to the case for about $200K in internal research funding. EY later showcased the project at the India AI Impact Summit 2026, after my departure.

Prototypes

Independent and academic builds.

Independent AI research and a group academic workflow prototype.

AI-assisted equity research

ArthLens

Turns an Indian company name, NSE symbol, or BSE code into a structured starting point for equity research and interview preparation.

Solo project · Research prototype

What I did: Built a solo, AI-assisted research platform combining public market data, financial calculations, peer comparisons, and industry context.
Result: Produces research reports with financial ratios, beta estimates, up to four rule-based peers, risks, catalysts, and verification links. Intended for preliminary research, not verified investment advice.

Open prototype

Multi-agent AI research

Concall Signal Agent

Explores how directly management answers earnings-call questions and compares subsequent stock performance with the NIFTY IT benchmark.

Full-stack research prototype

What I did: Built a five-agent analysis platform for 12 Indian IT-services calls, with batch processing, question-answer evidence, interactive visualizations, and blind human review.
Result: Validated the prototype with 22 automated backend/frontend tests. Deterministic scoring methods showed 0.998 correlation; this measures agreement between methods, not predictive investment accuracy.

Open prototype

Healthcare workflow prototype

DPVAS — Digital Prescription Verification

Explores antibiotic traceability through a shared workflow for doctors, patients, pharmacists, suppliers, and government auditors.

Group academic project · Simulated data

What I did: Contributed to a group academic prototype connecting prescription verification, quantity-limited dispensing, stock records, and audit visibility.
Result: Demonstrates an end-to-end workflow using fictional, browser-session data. No trained AI model, database, real SMS, or connection to healthcare or government systems.

Open prototype

Business

Problem structuring · Business analysis · Enterprise presales · Digital transformation · Stakeholder engagement

Technology

React and TypeScript · AI integration · Azure · Real-time systems · Robotics and NVIDIA