MartinMuskov

San Francisco, California · In-Person

Co-founder. Credda is an AI-driven software engineering platform that turns a bug report into a patch.

• Created the patch engine end to end: it replicates the reported error, localizes the root cause, generates the repair and validates the whole process automatically.
• Built the backend infrastructure and the artificial intelligence pieces that keep the platform running.
• Implemented the CLI, GitHub Action and client SDKs that customers use to run Credda inside their own CI workflows, so their source never leaves their environment.

San Francisco, California · Hybrid

The Data Fluency Initiative: the reporting and data tooling behind underwriting, actuarial, insurance operations and product.

• Co-developed the Converge BI underwriting dashboard, which brought scattered requests into one report. Designed the filters, views and chart types around what each team needed, then validated every figure in it against the actuarial team's independently calculated numbers and ran down the differences.
• Analysed the historical submission and policy data independently, which turned up two errors in the underlying data and in the forecasting model, both corrected before any conclusion was drawn. Delivered a written report on broker and segment performance against loss ratio and retention, with a premium forecast for the coming period.
• Improved the triage severity model, and established that its reliability is bounded by how few medium- and high-severity claims are labelled at all. Investigated ways to make the confidence an LLM reports about its own answer mean something.

Santa Cruz, California

B.S. Electrical Engineering, with a planned minor in Computer Science.