Building evidence-driven systems for
institutional digital asset risk.
I’m Arpit Pandey. I work in institutional financial operations and build
public systems that connect business intent with verifiable blockchain evidence —
across stablecoin settlement controls, protocol verification, blockchain finality,
market structure, and disciplined digital-asset research.
Research lensMarket structure, protocol verification & provenance
01 / Primary institutional system
Settlement controls built around evidence, not assumptions.
The control plane asks a simple institutional question: if a settlement
instruction says one thing and the blockchain records another, what can
we prove — and what must remain unknown?
01
Stablecoin settlement assurance
Institutional Digital Asset Control Plane
An engineering MVP for strict direct-transfer reconciliation on Base
Mainnet using canonical USDC reference data and auditable,
evidence-aware control outcomes. The current public build also includes
a Streamlit control-review dashboard with six synthetic settlement
scenarios and 70 automated tests (64 unit + 6 headless Streamlit
integration tests).
Modeled chain, asset, sender, receiver, amount and business asset identity.
02Observed evidence
RPC chain identity, receipt status, block identity and decoded ERC-20 transfer evidence.
03Independent controls
Each control evaluates only the evidence it actually requires.
04Audit bundle
Normalized findings plus detailed underlying results, without a premature business verdict.
Control coverage
Execution success is not settlement correctness.
A transaction can execute successfully while chain, asset, sender,
receiver, amount, finality, uniqueness, or canonical-asset controls
independently identify a problem.
These systems separate observation from interpretation, preserve failed
hypotheses and missing evidence, and keep conclusions inside what the
underlying data can actually support.
A cross-venue research system studying Kuru / Monad state-specific
order books alongside reconstructed Coinbase MON-USD Level2 evidence.
It uses point-in-time alignment, freshness controls, executable-depth
analysis, deterministic research outputs, and pre-registered methodologies.
Current verification: 297 tests passing.
The confirmatory dataset met the pre-registered descriptive association rule:
greater proposed-versus-finalized midpoint disagreement magnitude was
associated with greater subsequent absolute Coinbase midpoint movement.
No claim of causality, prediction, alpha, or arbitrage is made.
Event-level verification of 135 successful native-ETH WETH9 calls of
at least 1,000 ETH, separating selector evidence from verified protocol
execution and avoiding unsupported interpretation of trailing calldata.
A tested whale-flow research pipeline combining on-chain transfer
detection, USD normalization, liquidity and volatility context,
BTC benchmark-adjusted outcome validation, persistent research datasets,
backtesting, and a Streamlit analysis surface.
Current verification: 214 tests passing · 91% coverage · green CI.
Current evidence does not support positive ETH whale-flow as a standalone
durable signal after BTC benchmark adjustment. Failed signals are retained
as research evidence rather than rewritten as success.
Supporting work covers point-in-time data capture, provenance,
SHA-256 integrity checks, version-bound research artifacts,
exact raw-unit accounting, and explicit handling of unavailable evidence.
Institutional finance on one side. Verifiable on-chain evidence on the other.
I currently work in institutional financial operations at S&P Global.
My public work is focused on understanding the bridge between traditional
control logic and blockchain-based financial infrastructure.
I care about questions that sit between business intent and technical
evidence: what was supposed to settle, what actually happened on-chain,
which evidence is authoritative, how exceptions should be surfaced, and
where a system must refuse to make a claim.
I’m building a focused body of work around digital-asset settlement,
controls, operations, protocol verification, and evidence-driven research —
with enough technical depth for the underlying evidence to be reviewed.
This portfolio represents personal research and engineering work only.
It does not represent S&P Global, its clients, or any production
financial system.
Operating standard
Claims should survive review.
The standard across these projects is deliberately conservative:
preserve evidence, show limitations, and do not convert uncertainty into
a stronger claim than the data supports.
01
Public evidence only
No client data, production settlement data, credentials, or private operational information in public work.
02
No fake completeness
Unavailable evidence remains unavailable. Missing data is never silently replaced with synthetic facts.
03
Independent controls
A failure in one control does not rewrite the evidence produced by another independent control.
04
Explicit limitations
Engineering MVPs and research prototypes are described as such rather than presented as production systems.
Open work
Interested in institutional digital-asset risk, controls and on-chain infrastructure?
The repositories are public so the architecture, tests, assumptions,
limitations, and commit history can be reviewed directly.