About

I started as a chemical engineer, which means I learned early that real systems are constrained, messy, and punishing when you get the fundamentals wrong. That mindset — process thinking, first-principles analysis, respect for edge cases — is what I bring to every data problem I work on today.

Over five years of professional work I have moved through enterprise banking automation, production ML at scale, cross-border e-commerce analytics, and VP-level quantitative research. I now work on crypto HFT and MFT strategy development. Each stage compounded the previous one.

I am at my best when the problem is hard, the data is noisy, and someone needs to bridge the gap between research and production.

The Journey

Chapter 1 — The Foundation (IIT Bombay, 2017–2021)

I ranked AIR 1382 in JEE Advanced out of 200,000 candidates and AIR 735 in JEE Mains out of 1.3 million. Those numbers matter not because of prestige but because of what the preparation process taught me: systematic thinking under pressure, structured problem decomposition, and the discipline of working within constraints that cannot be negotiated away.

At IIT Bombay I studied chemical engineering. Batch processes, reaction equilibria, thermodynamics, fluid dynamics — these are not abstract subjects. They are operational systems with real failure modes. Engineering taught me to ask what breaks before asking what works.

In early 2019 I spent a semester at Danmarks Tekniske Universitet (DTU) in Copenhagen as an exchange scholar. I took lab courses in batch distillation, liquid-liquid extraction, gas flow in pipes, and filtration. These were hands-on experimental reports — the kind of work where the variance in your measurements is your problem to explain, not ignore. That semester shaped how I think about data quality: measurement error is a design decision, not a noise problem.

Late 2019 I joined Prof. Bryan Hooi’s lab at NUS for a two-month AI research internship. My first real exposure to machine learning research — specifically temporal graph learning and link prediction on dynamic networks. The College Messages dataset. AUC 86%. Small scope, but the mindset shift was permanent: models have assumptions, and assumptions have failure modes.

Key insight: Engineering taught me to ask what breaks before asking what works.

Chapter 2 — Learning How Real Systems Fail (ANZ Bank, Jun 2021–Mar 2022)

I joined ANZ Bank straight out of IIT Bombay. My first production role. The job was enterprise data automation — connecting regulatory reporting pipelines across APRA compliance requirements.

I built ETL workflows integrating 70+ source systems using IBM DataStage and Teradata. I worked with ASCII and EBCDIC file formats, mainframe-era data structures, and compliance requirements that had no tolerance for approximation. A wrong byte in a regulatory report is not a technical debt item — it is a legal event.

The most important outcome was cutting development and testing time by 50% through automation of previously manual validation workflows. The second most important outcome was understanding why enterprise systems are conservative: they are optimized for correctness at the cost of speed, not the other way around.

Key insight: In banking, correctness is not a nice-to-have. One wrong byte costs compliance.

Chapter 3 — Production ML at Scale (Blue Yonder, Apr 2022–Mar 2023)

Blue Yonder is a global supply-chain software company. I joined as a data scientist on a 15-person team working on production forecasting systems. The scale was real: 5TB+ of supply-chain data, seven distinct forecasting problem types, and a deployment pipeline that had to run reliably for enterprise clients.

I worked on fulfillment capacity prediction, delivery time estimation, replenishment optimization, markdown decision support, and inventory management. Each problem had different data distributions, different business stakes, and different latency requirements.

The technical stack was serious: TFX, Apache Beam, Google Dataflow, XGBoost, LightGBM, distributed training across GCP. I learned the difference between a model that scores 95% on a held-out test set and a model that runs in under 30 seconds at 3AM in a client’s production cluster. They are different products.

Key insight: A model that’s 95% accurate but can’t run in 30 seconds in production is worthless.

Chapter 4 — The Consulting Turn (GoGlocal, Mar–Oct 2023)

GoGlocal is a cross-border e-commerce intelligence company. I joined as Manager of Data Science — my first role owning the full data product rather than a component of someone else’s system.

The scope: 1,000+ SKUs across Amazon, eBay, Walmart, and Lazada marketplaces. I built NLP pipelines for product classification and attribute extraction, pricing intelligence systems, competitor analysis automation, and revenue estimation models. I reduced manual effort across the team by 50% and improved revenue efficiency by 30%.

The consulting dimension meant constant translation between technical outputs and business decisions. A pricing model that a data scientist is proud of is worthless if the merchandising team cannot act on it. I learned to frame model outputs in terms of decisions, not metrics.

Key insight: Real business value means translating model output into a decision someone can act on.

Chapter 5 — Going Deep on Markets (Mastertrust, Jan 2024–2026)

Mastertrust is a broker-dealer and trading firm. I joined as VP of Data Science and quantitative research — the most demanding role of my career.

I owned quantitative research, machine learning systems, and production trading infrastructure simultaneously. The flagship outcome: Sharpe ratio of 4 in live systematic index options strategies. I managed portfolios exceeding ₹100 crore (~$12M USD) in live capital.

The technical work was deep: volatility surface modeling, microstructure analysis, regime detection, execution-aware backtesting, walk-forward validation, options Greeks as risk management instruments. The business work was equally demanding: communicating strategy logic to senior stakeholders, managing capital allocation decisions, and maintaining the discipline to walk away from strategies that looked good on paper but had fragile assumptions.

Markets are the most honest feedback loop I have ever worked in. They punish overfitting instantly and without mercy.

Key insight: Markets punish overfitting instantly and without mercy. That feedback loop is unlike anything else.

Chapter 6 — Current Work (NK Securities, 2026–Present)

I now work at NK Securities India on crypto HFT and MFT systematic strategy research.

Crypto markets have structural properties that make them uniquely interesting: 24/7 operation, fragmented liquidity across exchanges, high variance in microstructure quality, and adversarial order flow that rewards execution discipline as much as signal quality.

I am applying microstructure analysis to high-frequency crypto order flow — building execution-aware models where slippage, latency, and position sizing matter as much as signal alpha.

Chapter 7 — The Through-Line

Chemical engineering → enterprise automation → production ML → quantitative research.

Every stage compounded. Process thinking from engineering transferred to pipeline design. Correctness discipline from banking transferred to production model reliability. Scale thinking from supply-chain ML transferred to backtesting rigor. Business translation skills from consulting transferred to communicating trading strategy logic to stakeholders.

I am a rare hybrid: capable of business advisory and deep technical research in the same conversation. I can read a volatility surface, write a production pipeline, and explain both to a client who has never opened a Jupyter notebook.

That combination is what I offer.

How I Think

First principles before frameworks. Every framework is someone else’s abstraction over a problem they had. Before I reach for a library or a methodology, I want to understand why the underlying problem is shaped the way it is.

Systems thinking from engineering. Every production system has inputs, outputs, failure modes, and feedback loops. I model data systems the same way I modeled chemical processes — with attention to what happens at the boundary conditions, not just the expected case.

Proof before confidence. A claim without a metric is a hypothesis. I try to distinguish clearly between what I know, what I believe, and what I am guessing. The career proof ledger below reflects this.

Write it down — if you can’t explain it, you don’t understand it. I have maintained a writing practice across ML, quant, data engineering, and industry topics. Writing is how I verify my own understanding, not how I display it.

Production is the only real test. Research that never ships is interesting. Research that ships and holds up is rare. My track record is weighted toward things that actually ran in production.

Selected Proof

Outcome	Context
Sharpe 4 in live index options	Mastertrust systematic trading
₹100 crore+ portfolio managed	Mastertrust, live capital
70+ source systems integrated	ANZ Bank APRA compliance
50% dev/test time reduction	ANZ Bank automation
5TB+ supply-chain data processed	Blue Yonder forecasting platform
1,000+ SKUs automated	GoGlocal e-commerce intelligence
50% manual effort reduction	GoGlocal
86% AUC — graph link prediction	NUS AI Research, Prof. Bryan Hooi
JEE Advanced AIR 1382 / 200,000	IIT Bombay admission
JEE Mains AIR 735 / 1,300,000	Top 0.06%

Education

Indian Institute of Technology Bombay B.Tech Chemical Engineering · May 2021 · CGPA 7.7

Danmarks Tekniske Universitet (DTU Copenhagen) Exchange Scholar · Jan–Jun 2019 Lab courses: Batch Distillation, Liquid-Liquid Extraction, Gas Flow in Pipes, Filtration Experimental reports available in the Credentials appendix.

What I’m Looking For

I take select consulting engagements in quantitative research, machine learning systems, and data automation. I am also open to senior research and applied AI roles at firms where the problems are genuinely hard.

If you want to work together, the best first step is a direct conversation.

Career Journey