A Conversation with Harshil Suvarnkar | Aditya Birla Sun Life AMC Ltd. | Young Manager Series S01E02

ELI5/TLDR

Harshil Suvarnkar spent a decade managing bonds and interest-rate risk from a treasurer’s chair, then switched to managing money for Aditya Birla Sun Life. His unusual edge: he understands both sides of the bond market (who borrows and who invests), which teaches you that duration isn’t just about macro signals—it’s about who’s actually buying and selling right now. He uses regression models to stress-test his macro hunches, treats AI as a pattern-completion tool (not an oracle), and learned from the 2013 rate shock that conviction matters more than half-measures.

The Full Story

A Dual-Sided Education

Harshil didn’t drift into fixed income by accident. He grew up managing his family’s finances, studied commerce at Sydenham College, then picked up an MBA and a diploma in securities law from JBIMS. The pull toward investing clicked when he realized it rewards relentless learning—you have to know GDP, fiscal policy, commodity cycles, credit quality, all at once. He interned at HSBC building a peer-review model for banks; years later, HSBC called asking for that same model. A small signal, but it showed him that rigor compounds.

His tenure at a major housing finance company proved more useful than most MBAs. He started in the treasury, managing liabilities and borrowing programs. Then he noticed the company’s investment side was suboptimal, optimized it, and got merged into the role. Suddenly he was straddling both sides of the ledger: raising capital (floating-rate liability) and investing it (credit selection, duration calls). He learned to hedge using interest-rate swaps and OIS markets. He interacted with investment managers, banks, and law firms. That dual exposure is rare. When he moved to Aditya Birla Sun Life to manage fixed-income portfolios on the buy side, the transition was smooth—he already knew the ecosystem from the other angle.

“If you do borrowing and investments, you interact on the buy side, you interact on the sell side, and you interact with intermediaries. You get exposure to what happens in the buy side also because the buy side services you.”

The Macro-Micro Lattice

Standard fixed-income wisdom says: watch duration, credit quality, yield curves. Harshil doesn’t disagree, but he stacks layers. Duration isn’t binary (up or down)—it’s a convex macro risk whose payoff depends on where you’re positioned, what inflation does, what fiscal and liquidity do. Getting the macro call right is necessary but not sufficient. You also need the micro call: demand-supply dynamics at the granular level.

The Reserve Bank buying and selling is the elephant in the room. But RBI’s actions depend on external shocks like oil and FX, which trigger cascading liquidity effects. Even more granular: state governments borrow to fund populist schemes, flooding the state development loan market. If you own that paper, you can’t exit without taking spread losses. A yield curve barbell (1-year and 10-year bonds) can perform very differently from a plain 3-5 year bond—not by chance, but because the curve itself has a shape that rewards certain positions. Knowing whether the 10-year or the 5-year will outperform is often the difference between winners and losers.

India’s macro depends on oil and currency (its two ruling variables) more than domestic factors alone. Oil shocks ripple through the balance of payments, which triggers RBI intervention and currency pressure, which then affects liquidity and yields. So he looks at India micro (which bond, which state, which issuer demand is soft or tight) but global macro (oil, geopolitics, US policy).

“You need to align multiple lenses together and then think through that one lens which transpires into your strategy.”

Equity Through a Debt Lens

Recently, Harshil started managing equity funds alongside his fixed-income work. Rather than abandon his macro framework, he weaponized it. Fixed income is inherently macro-driven, and rates are the discount rate in every DCF model. If you get the rate call wrong, your valuation is wrong. So he and his colleague do top-down (macro to sector, rates to capex cycle to commodity cycle) and then bottom-up (which companies in that sector). The edge: he knows management quality from years of interacting with issuers (they raise debt, he scrutinized them). He also understands REITs and InvITs better than pure equity managers because those asset classes behave like bonds—rate-sensitive, yield-driven.

Regression as a Mirror

When he asks “what is the sensitivity of yields to a $10 oil shock?”, he can’t rely on intuition alone in a market moving by microsecond. He built multiple regression models with GDP, fiscal, inflation, liquidity as inputs and yields as output. The models don’t predict the future (they can’t, because regimes break), but they let him ask: “In the past, when oil rose X%, yields rose Y%. Today, oil rose X%. My prior says yields should rise Y%, but they only rose Z%. Why the discrepancy?” That discrepancy is often signal.

He’s careful to exclude crisis periods from the regression (COVID when oil went negative, for instance) to strip out noise. He also thinks in Bayesian terms: start with priors (what happened historically), update them when new information arrives (geopolitical shock, policy shift), and adjust position size or direction accordingly. This is not mechanical—it’s using history as a compass, not a map.

AI and quant tools fit here as aids, not oracles. LLMs are pattern-completion machines. They’re excellent at summary, analogy, and context retrieval but weak at arithmetic and spatial reasoning. He uses AI for language-based research synthesis (connecting ideas across sources) but not for numbers. The internal team has built machine learning models that flag when portfolio decisions have historically led to certain outcomes, helping refine the decision process—but again, the tools inform; they don’t decide.

“They are to be used as aids in decision-making process. Given that you have more structured data and open data availability now, you can use these tools to take a better decision. If you don’t use that, somebody else will.”

The 2013 Lesson: Conviction Without Equivocation

In 2013, three years into his investing career, rates were low but inflation was rising and the balance of payments was deteriorating. He had conviction that RBI would hike, and it did—by 320 basis points in one cycle. Here’s where his learning arrived: he exited 60% of the portfolio. In hindsight, if he truly believed the thesis, he should have reduced duration across the board instead of half-exiting. Loss aversion—the psychological sting of marking losses on paper—made him timid even though he was right.

His boss supported him that morning (called at 8 AM with “it’s going to be a frenzy”; got approval at 9:01 to sell). But the lesson stuck: if you have conviction and get the call right, take it fully. Don’t hedge it with a half-measure because you’re afraid of the short-term pain. That year, traders joked about “helmet time”—going to the office in armor. It was brutal. But it taught him to know his biases (loss aversion, recency, anchoring) and build processes that favor System 2 thinking (slow, deliberate, analytical) over System 1 (fast, emotional, reactionary).

Mental Models and Reading

He’s influenced by philosophy as much as finance. He does solo skydiving—his trainer taught him that fear is useful because it makes you check your gear. The analogy carries: in investing, healthy fear makes you stress-test assumptions, size positions carefully, and have exit plans. He frequents ISKCON and finds the Bhagavad Gita instructive: written in the midst of a crisis (the Kurukshetra War), it teaches nishkam bhava (detachment from outcomes). Investing, he argues, is a journey. If you fixate on hitting a numerical target, you’ll be biased in your decisions. Better to optimize your process, do your best analysis, and accept that returns flow from discipline and intellectual humility.

His canon includes Daniel Kahneman’s Thinking, Fast and Slow (System 1 vs. System 2), Pulak Prasad’s What I Learned About Investing from Darwin (survival of the fittest applied to company selection; focus on ROCE), Michael Mauboussin’s work on expectations and Bayesian thinking (probabilistic forecasting, conditional priors), and Manish Dangi’s Booms, Busts and Market Cycles (India-rooted frameworks for understanding cycles). He reads 30-45 minutes before sleep every night, treating it as compound interest on years of experience.

The Daily Ritual

His calendar is pedagogical. One hour before market open, he reads sell-side notes and news to understand what might move the market. Commute time, he’s reading or reviewing notes. Half an hour to 45 minutes to fitness (walk or gym). Family time. Another 30-45 minutes of reading before bed. He’s engineered his information diet: he follows specific thought leaders on Twitter rather than accepting algorithmic feeds, because algorithmic curation amplifies bias (it shows you what you already believe). He avoids social media distraction because the cost of being algorithmically nudged is higher than the benefit of breaking news.

Key Takeaways

Dual experience beats single perspective. Having worked both the liability side (raising capital, managing duration risk) and the asset side (investing, credit selection), his mental models are richer and less vulnerable to the blind spots of pure buy-side managers.
Macro and micro are inseparable in fixed income. Global oil and currency matter for India’s yields, but so does hyperlocal state-government borrowing. Both scales demand attention.
Regression is a mirror, not a fortune teller. Quantitative models help you notice when the market is behaving differently from the past. The discrepancy is often the trade.
AI is pattern completion, not logic. Use LLMs for synthesis and idea connection. For arithmetic, numerical forecasting, and logical inference, you’ll lose money trusting the tool.
Loss aversion kills conviction. If you have a thesis and the facts support it, the half-measure (exiting 60% instead of acting fully) is a behavioral trap. Know your biases.
Reading compounds faster than experience. A 25-year career in 400 pages beats learning only from your own trades.
System 2 thinking demands process design. You can’t just “be deliberate” in a crisis. Build workflows that force slow, analytical decision-making at key checkpoints.

Claude’s Take

This is genuinely thoughtful investing. Harshil doesn’t pretend bonds are stocks, doesn’t overstate what AI can do, and doesn’t hide behind pseudoscientific backtesting. His strength is integrating multiple layers of analysis—macro regime, micro demand, quantitative sanity checks, behavioral guardrails—without pretending they collapse into a single number. The 2013 anecdote is the most valuable 15 minutes of the conversation: it’s not “I was right and made money”; it’s “I was half-right but psychologically half-exited and learned something about myself I’ve had to work on for a decade.”

The weakness, if there is one, is that his framework is hard to systematize for a team. It relies on judgment calls (which state will borrow, which curve will outperform) and an almost omnivorous curiosity. Not every analyst will read Kahneman at night or meditate on the Gita. The philosophy is sound, but the repeatability is low. For a manager running billions, that’s a real limitation. Still, for someone trying to think clearly about debt markets and their own blindspots, this is a master class. The mentorship-style delivery and frank discussion of his own mistakes elevate it well above the typical “I crushed it” fund manager podcast.

Credibility is high. He’s candid about AI’s limitations, avoids the hype cycle, and grounds everything in first-principles reasoning. A skeptical finance person won’t feel lectured. A young analyst should find multiple ideas to steal. Score reflects the density and usability of the insights, not the charisma of the speaker.