const arun = {
  role: "Founding AI Engineer @ Lamatic.ai",
  focus: ["agent runtimes", "multi-agent orchestration", "production LLM ops"],
  building: "infrastructure for durable autonomous systems — not slide-deck demos",
  firstHire: true,          // shipped 80%+ of the core platform
  stack: ["Python", "TypeScript", "Go", "Cloudflare Workers", "Kubernetes"],
  caresAbout: ["durability", "distributed execution", "eval", "what breaks after the millionth run"],
};

I'm Arun Addagatla, an AI systems engineer who builds the runtime glue that keeps agents alive past a single HTTP request — serverless execution, multi-agent orchestration, memory/RAG layers, and eval. As the first engineering hire at Lamatic.ai (TechCrunch Startup Battlefield 200 · Cloudflare Workers Launchpad), I've shipped over 80% of the platform.

A multi-agent harness that turns a plain-English request into a validated, deployable Lamatic workflow. I first built a reusable multi-agent node primitive, then dogfooded it to build the harness itself.

flowchart LR
    P[🧭 Planner] --> A[🏗️ Architect]
    A --> W[🧵 Weaver]
    W <--> I[🔍 Inspector]
    I --> F[🧩 Filler]
    F <--> Au[🛡️ Auditor]
    Au --> V[✅ Validator]
    V --> O[(Deployable Flow)]

Two-tier verification: deterministic checks where ground truth exists; a rubric LLM-as-a-judge only for fuzzy output quality — and the judge never overrides a hard check.

AI content-orchestration monorepo guiding the full pipeline — clarify → outline → write → edit → publish — with autonomous agents owning each stage.

An MCP server exposing 35+ tools over the Dev.to (Forem) API. Lets Claude & Cursor draft, edit, publish, and manage content. Supports stdio · Streamable HTTP · Cloudflare Workers transports, with a multi-arch Docker image on GHCR.

Local CLI that searches jobs, dedupes across runs, and drafts structured resumes + cover letters from your base profile using LangGraph agents — with checkpointing, optional PDF export, and pluggable models (Gemini, Anthropic, OpenAI, Ollama).

Hands-on collection of LLM & agent experiments — implementations and deep-dives that back my writing on memory, RAG, MCP, and multi-agent systems.

End-to-end voice assistant across 10+ Indian languages. Conformer S2T on Triton, Fastpitch TTS, and a LangChain RAG pipeline with embedding + reranker models.

Lamatic.ai — Founding Engineer, AI · Mar 2024 – Present · Miami, FL (remote)

• Built 80%+ of the core stack as the first engineering hire — AI systems, backend, infra, and critical frontend.
• Architected a serverless executor at 1M+ monthly runs and a deployment engine at 1K+ deploys/min.
• Cut deployment latency 2 min → 15 s (~87%).
• Built the flow-gen multi-agent harness and a configurable LLM-as-a-judge eval framework.
• Shipped an internal hiring agent (resume parsing + video analysis) cutting recruiter workload 70%.
• Engineered secure VPC Kubernetes ETL for Drive/S3/SharePoint with OAuth, Semantic RAG, a Multi-Agent Supervisor, Slack/Teams webhooks, and a native GitHub VCS for flow sync.

Samespace — SDE, AI/ML · Oct 2022 – Mar 2024 · Mumbai

• Chat/voicebots handling 95% of queries (embeddings, reranking, tuned Zephyr & GPT-4).
• Fine-tuned Mistral / LLaMA-2 with LoRA/PEFT (+30% fluency); LLM inference engine at 106 tok/s.
• Optimized Whisper V3 with ONNX/TensorRT + Triton → 0.1–0.4s latency; multimodal search (+60%).

Enterpret — ML Intern, NLP · Sep 2021 – Aug 2022 · Bangalore

• Serverless multilingual sentiment on AWS (−50% processing time); CI/CD; NER/classification (+30%); anomaly detection (−60% false positives).

Algoritmo Labs — Data Science Intern · 2020 — ML model → ONNX → Go runtime for client-side deployment.

Languages

Web & Backend

AI / LLM / ML

Inference & Model Serving

Data & Vector Stores

Cloud & Infra

Observability & CI/CD

Recent talks

• Why LLMs Need Memory — Lamatic Community (Mar 2026) · YouTube
• Applications of AI — Omkaranada Institute (Apr 2026) · live-shipped a web app in <8 min
• Why Prompting Isn't Enough: The Case for RAG — Lamatic Community (Jan 2026)
• What is MCP & How It Works — Daytona Developers Club Tour '25, Mumbai (May 2025)

Recent writing (70k+ reads across Medium · GoPenAI · Nerd For Tech · Lamatic Labs)

• Why LLMs Need Memory — Building AI Agents Hands-On · Mar 2026
• Inside NVIDIA Nemotron 3: Hybrid MoE for Multi-Agent AI · Dec 2025
• Cut Token Costs by 60%: TOON vs JSON for AI Workflows · Nov 2025
• Lamatic — The Operating System for AI Agents

Building reliable GenAI + Agentic AI for enterprise · durable execution/inference platforms · production MLOps · RAG, MCP integrations, and autonomous workflow automation.

📊 GitHub

✨ "Build useful AI. Ship it fast. Scale it responsibly." ✨