Vyshnavi DP

AI and full-stack engineer with 6+ years across Dell Technologies and TCS, and an M.S. Software Engineering from San Jose State (May 2026). At Dell I was embedded in the SupportAssist product team (via Accenture). Right now I work on AI agent workflows and the evals that measure them — agent orchestration, tool calling, and regression-gated LLM evaluation — built on full-stack production engineering.

The thread across eight years is trust in automated systems: retail optimization where merchandisers needed explanations and overrides before acting on recommendations; device automation on Dell's preinstalled PC platform, where a remediation script is remote code execution unless the signature chain holds and a false-positive failure prediction ships a physical replacement part; and agent systems evaluated on tool-call accuracy, schema adherence, and error recovery, with failure taxonomies driving redesign.

My open-source work spans an AI code review evaluator (Sentinel), a multi-tenant OKR tracker with database-enforced tenant isolation (Kairos), a Go time-series database (Helios), adversarial robustness research (NeuroLens), and an autonomous agent with a statistical eval harness (Archon). The thread across all of them: ship the production pipeline and the measurement harness in the same repo.