rudra Trivedi

With Lifed, I engineered a local-first personal AI command center that combines local Gemma and Llama 3 agents with Gemini 2.0 Flash, and can fall back to fully offline operation. Its tool-calling loop executes tools against SQLite while keeping private data on the device.

I implemented on-device semantic memory with FastEmbed and ONNX Runtime, plus a planning engine that separates deterministic task ranking from LLM-generated explanations. I also built a voice pipeline with an offline fallback and tested the system’s vector math, tool schemas, and planning heuristics.

For my Voice-First RAG Research Assistant, I built a citation-grounded question-answering system for uploaded documents, with guardrails for cases where retrieved context is absent. Its retrieval pipeline combines Qdrant dense search with cross-encoder reranking, and supports local Ollama and Groq Cloud backends.

I also built an Enterprise AI Data Analyst Platform with a five-agent workflow for data cleaning, statistical analysis, forecasting, and visualization. I’m seeking a remote AI Engineer role.