Vyshnavi DP
AI & Full-Stack Engineer · San Jose, CA
Education
Experience
- ▸Built C#/.NET microservices for Dell SupportAssist — diagnostics, driver/BIOS update delivery with signature verification, and telemetry for a platform with 10M+ installs (my services processing 2M+ active devices)
- ▸Re-architected scan status from polling to WebSocket push, cutting steady-state load ~60%
- ▸Built the in-app AI Virtual Assistant experience with device-context injection and assistant-triggered actions
- ▸Owned full SDLC for 4 major features: technical design → dev → QA → production rollout
- ▸Mentored 3 junior engineers; introduced PR review standards and component documentation practices
- ▸Developed enterprise retail analytics SaaS for Fortune 500 clients using Angular 9 and Spring Boot
- ▸Implemented virtual scrolling for 100K+ SKU inventory tables; reduced response time from 800ms to 120ms
- ▸Built Redis-backed caching layer for frequently accessed analytics dashboards (80% cache hit rate)
- ▸Integrated Spring Batch for nightly ETL pipelines processing 50M+ retail transactions
- ▸Delivered REST APIs for client-facing analytics endpoints used by 3 enterprise retailers
Projects
Autonomous agent in Python — planner/executor/reflector with a typed middleware chain (schema validation, retries, tracing on every tool call) and a statistical evaluation harness: seeded runs, per-step traces, and a failure-mode taxonomy.
GitHub App that reviews PRs via BM25 + dense (pgvector) retrieval, cost guardrails, and a 98-fixture eval harness — per-category P/R/F1, regression-gated in CI.
Multi-tenant OKR tracker that enforces tenant isolation in PostgreSQL with Row-Level Security, not in application code.
WAL + LSM storage, Gorilla compression, label-postings filtering, a PromQL subset, streaming EWMA anomaly detection, and optional 3-node Raft replication.
MVP adversarial-ML research: from-scratch PyTorch models (no torchvision pretrained), FGSM and PGD attacks implemented from the original papers, defenses, and a multimodal-to-unimodal attack-transfer experiment.