Systems read
Cross-domain
Open threads

I build where software meets physics, markets, institutions, and open research questions.

Incoming at Tesla on finance, AI, and large-scale systems. In parallel I am in avionics at Skyroot on safety-critical AI and traceability, finishing an M.Tech. in Software Reliability (QRE) at IIT Kharagpur, and keeping peer-reviewed research on the same calendar as shipping work across GovTech and early startups.

Before that: LinkedIn, Scale AI, Samagra, IISc, and Mitacs—from huge data pipelines to graph work, misinformation studies, and open digital public goods. More story and papers live on Profile and the publication list on Corpus.

Ansh Sarkar
Institutions
Bench + scale
Marquee stream
TeslaSkyrootIIT KharagpurLinkedInScale AISamagraIIScMitacsC4GTCNCF
Reliability
Research receipts
Shipped surfaces

Three rails that stay in tension on purpose.

Reliability as a lens

Treat tail risk, evaluation, and operator-grade UX as one system. QRE coursework and production scars share the same vocabulary.

Research with receipts

Peer-reviewed work spanning game theory, cryptography, NLP pipelines, and data engineering. Corpus stays in sync with the public record.

Shipped interfaces

From GovTech microservices to avionics benches: fewer hand-wavy diagrams, more measurable behavior under real constraints.

Constraint-first
Multi-sector
Names, no marks

How the same mind holds many tools.

Not a scatter plot of hobbies, but a single obsession with getting complex systems to behave under real constraints: latency, regulation, capital, and human attention.

AI & ML systems

Shipping models, data planes, and evaluation loops in production contexts.

Finance & quant

WorldQuant BRAIN (Gold), algorithmic trading coursework, pricing & risk intuition.

Reliability & safety

Formal reliability engineering, testing philosophy, and large-scale correctness.

GovTech & platforms

Samagra, C4GT, DMP mentoring: population-scale digital public infrastructure.

Research

Peer-reviewed threads from game theory and cryptography through NLP and data pipelines; IISc and Mitacs fieldwork; questions about evaluation, graphs, and reliability of modern AI stacks.

Aerospace

Skyroot avionics: signal integrity, verification, launch-adjacent software paths.

Owned markdown
Feed mirror
Long-form bias

From the bench

All posts & feeds →

Longer essays live here; shorter notes also show up from Substack and Medium on the journal page. ✍️ Projects is where the really big programs get their own pages—timelines, links, and context that does not fit in one blog post. 🏛️