Hi 👋, I'm Anirudh Lath

Software Engineer III @ Walmart (VIZIO SmartCast) — Generative AI & Agentic AI

I build production LLM systems: autonomous tool-calling agents, Retrieval-Augmented Generation (RAG) with vector search, real-time LLM guardrails, and evaluation infrastructure for an in-house voice assistant on VIZIO Smart TVs.

What I work on

• 🤖 Agentic AI — server-side tool registries, structured outputs, multi-step tool-calling agents (Vertex AI / Gemini)
• 🛡️ LLM safety & guardrails — semantic guardrail engines (embeddings + vector search + LLM-judge fallback), adversarial red-teaming, refusal attribution
• 📊 LLM evaluation — eval platforms, A/B harnesses, LLM-as-judge scoring, eval-driven development
• ⚙️ Systems & platform — C++17 declarative UI engine, Rust language tooling (LSP), FastAPI/WebSocket backends, Azure
• 📄 Patent pending: context-based input interfaces for smart TVs (US 2025/0231778 A1)

Featured projects

• alfred — local-first multi-agent voice assistant: dual-process architecture (local Ollama SLM fast path + Claude agentic tool-use loop), LLM-generated automation triggers, three-layer memory with embedding search, DeepEval evaluation suite. Python, FastAPI, Redis, MQTT — with a companion SwiftUI iOS client and Home Assistant microservice.
• meta-mcp — Model Context Protocol (MCP) router that selects tools using embeddings, semantic search, and RAG over a Qdrant vector database. Python, FastAPI, Qdrant, Docker.
• moneta — personal-finance engine where the LLM is a classifier gating decisions, never a source of money values: LLM-gated recurring detection with a human review queue, RSU-vesting-aware net worth, deferred-interest warnings. Python 3.13, SQLAlchemy async, LiteLLM.
• taxscan — sales-tax deduction analyzer: Gmail receipt extraction, LLM classification with enforced structured outputs (json_schema + $defs inlining), fuzzy receipt-transaction matching. Python 3.13, SQLModel, LiteLLM.
• dj-ledfx — beat-synced LED lighting engine driven by Pioneer Pro DJ Link: 60fps effect renderer with per-device latency compensation, Prometheus/Grafana observability. Python, FastAPI, React.
• genai-image-studio — DreamBooth fine-tuning and image-generation web service. PyTorch, FastAPI, Gradio, Docker.
• AsyncCardStack — modern async/await SwiftUI card-stack library.

Reach me

• 💼 linkedin.com/in/anirudhlath
• 📫 anirudhlath@outlook.com
• ⚡ Fun fact: I'm a filmmaker and an artist too!