KM Kanav Mehta Quantitative Researcher

QuantLink AI · Agent Platform

A quant desk, opened to AI agents.

AI agents get a quant researcher's toolkit: 38 MCP tools run AutoML, DAG pipelines, backtests and risk attribution behind a blocking leakage and overfitting gate. A governed data server gives agents and notebook kernels permission-checked access to 789 datasets, returning large results as Parquet handles instead of flooding the agent's context. Built as lead engineer; verified end to end on staging.

More QuantLink work: portfolio construction, optimization and attribution, in production with 10+ college funds piloting · point-in-time data platform · notebook agent · notebook dashboards · AutoML workbench · operator catalog and DAG builder.

Back to Projects

Agent tooling

The desk's machinery, one tool call away.

  • 38 agent tools cover AutoML runs, DAG execution, a blocking leakage and overfit evaluation gate, feature-store queries, knowledge-graph search, DCF valuation, and notebook operations.
  • A stdio server connects the full platform to Claude Code or Cursor in one command; its HTTP twin shares the same MCP contract.
  • The data server lets an agent search the catalog, inspect schemas, preview a request, look up valid parameter values, fetch, and query in SQL across 789 datasets.

Governed data

Every byte authenticated, every dataset entitled.

  • Fail-closed by construction: with no token verifier installed, the MCP endpoint falls back to the API-key gate rather than opening up, and caller identity always comes from the verified token.
  • Dataset-fetching routes sit behind JWT, with fail-closed per-dataset access checks on fetch, preview, query, and spec compilation; the same gate covers the SQL relay and the notebook-kernel data broker.
  • Large results never flood an agent's context: each fetch is written once to Parquet under a deterministic query key, the agent gets a bounded preview plus a handle, and follow-up DuckDB SQL or an Arrow stream reads the snapshot without another provider call. Tool errors pass through structure-aware secret redaction, pinned by a catalog-wide test.

Speed & reliability

Fast enough for an agent loop, proven on staging.

  • A shared rate limiter and cache cut a parameter lookup from 1.35 s to 0.16 s, so agents can explore valid inputs without metered spend.
  • Notebook kernels reach the server with a scoped, silently refreshed grant, so credentials never appear in cell output.
  • Continuity lives in model-visible handles passed as tool arguments, the pattern the stateless MCP revision standardizes. Verified end to end on staging with an MCP SDK client: auth, fetch, query by handle, and Arrow read-back.