Saurav Sharma
Independent researcher and technical builder working on sequential decision-making, financial AI, and reliable action under hidden mechanisms.
- AffiliationIndependent Researcher
- EducationBS student, IIT Madras
- BackgroundFinancial markets & trading
From economic agents to a research program.
My current program began from attempts to build economic decision agents and evolved into a broader investigation of where capable systems fail when they must identify governing structure before acting, acquire information without destroying future options, or respect constraints that become decisive only under stress.
I work independently and use controlled environments, exact solvers where possible, frozen evaluations, adversarial baselines, and explicit negative results to separate mechanism claims from benchmark narratives.
Background.
Experience studying and trading financial markets shaped my interest in execution, liquidity, credit constraints, tail risk, and decision-making under uncertainty.
IIT Madras.
BS student, Indian Institute of Technology Madras.
Research affiliation remains Independent Researcher.
Current focus.
- timing information acquisition when epistemic actions have endogenous consequences;
- decision-critical model misspecification in financial constraints;
- hidden-rule and rare-event decision environments;
- external validation protocols for the broader AEI program.
Public evidence boundary.
The public portfolio is intentionally narrower than the full private archive. I prefer to expose enough for serious technical scrutiny while preserving implementation details that are not necessary to evaluate the scientific claim.
How the pieces connect.
Decision problems, not only prediction
Hands-on market experience and early agent experiments focused attention on feasibility, execution, tail risk, and governing rules.
Mechanism identification
Controlled economic worlds exposed a gap between knowing what to do under an explicit rule and identifying which rule governs the decision.
REC, private credit, and microstructure
Separate papers isolate timing information acquisition, boundary-moving model error, and endogenous value of information.
External validation of the broader thesis
The next research priority is externally specified evaluation rather than another internally designed success.
Contact & profiles
Research discussion · Reproduction · Collaboration