Skip to content
View NehharShah's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report NehharShah

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
NehharShah/README.md

Nihar Shah

I build software that acts.

I'm a systems engineer and researcher working across AI infrastructure, backend systems, developer tooling, and autonomous software.

At Ostium, I build low-latency infrastructure for automated hedging, risk, execution, market connectivity, and production reliability. I also built the internal agent platform used for engineering and operational workflows.

Previously at Subconscious AI, StarkWare, and Nethermind, I worked on research agents, model infrastructure, distributed systems, and verification.

I'm interested in the point where software stops producing answers and starts taking actions: understanding state, making decisions, interacting with real systems, and remaining observable, constrained, and recoverable.

Open source

dfm
A local coding agent for Apple Silicon that can inspect, edit, build, test, repair, and verify code using on-device models.

repokg
Builds an evidence-backed knowledge graph of a codebase: architecture, dependencies, repository history, operational surface, and structural change.

repocanon
Turns a repository into consistent, repo-specific context for Codex, Claude Code, Copilot, and Cursor.

Reflex · contributor
Local infrastructure for learning reusable workflows from coding-agent histories.

I also contribute to projects across developer infrastructure, databases, agent systems, and execution tooling.

Research

I research what changes when software can take consequential actions under uncertainty.

My work spans agent oversight, reliability and control, verification, mechanism design, market structure, and risk.

Recent work includes autonomous market making, liquidation and solvency, funding mechanisms, settlement verification, and tail-risk control for side-effecting agents.

My paper, Calibrated agents as market makers: automated liquidity provision for long-tail prediction markets, was recently accepted for publication in Digital Finance (Springer). The work establishes optimal market-making spread decompositions based on agent estimation error to solve liquidity constraints in sparse or uncertain environments.

Research →

Elsewhere

Website · Google Scholar · X · LinkedIn · XERJ

Pinned Loading

  1. yield-pool-watcher yield-pool-watcher Public

    Track APY and TVL across pools and alert on changes

    TypeScript

  2. repokg repokg Public

    repokg: AI ready knowledge graph of any codebase: modules, import edges, branch/PR state, ops surface. One command, zero deps.

    Python 3

  3. calibrated-propamm-standard calibrated-propamm-standard Public

    Rust

  4. sumervaid/reflex-tui sumervaid/reflex-tui Public

    Turns repeated agent sessions into skills and memory, and routes each task to a model by past sessions.

    TypeScript 8