Karthik Nadar

· Applied AI Engineer · figuring out stuff

Curious enough to break things. Stubborn enough to rebuild them better, more curious than qualified, and that’s the point.

Links: [LinkedIn] [Twitter] [GitHub] [ Resume]

Current Research Focus

I'm deeply interested in self-evolving agents and long-term AI memory, systems that don’t just respond, but accumulate experience, adapt their internal representations, and reshape themselves over time.

Inspired by large-scale multi-agent simulations like Stanford’s generative agents Smallville work, I’m curious whether artificial agents, when given memory, causality, and autonomy, might converge toward entirely different social structures, values, or evolutionary paths than humans.

If intelligence is substrate-agnostic, what kinds of worlds would it build without human priors?

Experience

• RapidNode (Remote Intern) — June 2025 - August 2025. Built ingestion pipelines to process millions of Discord and Twitter data points. Developed social graphs and dashboards.
• YesItaly (Intern) — August 2025 - November 2025. Built an in-house CRM from the ground up. Integrated Tata Tele's SmartFlo API to implement click-to-call and call operations. Designed an email dashboard to track incoming emails in real time and connected to the DB ads account for lead tracking.
• Brio Labs (Founding AI Engineer) — December 2025 - March 2026. Building the memory layer and knowledge graphs for data retrieved from Google Drive, OneDrive, and Dropbox. Developed retrieval pipelines that reduced ingestion time from 15 minutes to 4 minutes per file. Contributing to search engine development and improving prompting techniques through research.
• Nura (Founding engineer) - March 2026 - Present. Built and operated distributed web scrapers on AWS Lightsail to ingest publicly available county-level data at scale. Implemented cited answers for zoning ordinance queries, ensuring every response is grounded in source documents. Integrated the Google Routes API to compute accurate, real-time point-to-point distances for ordinance-driven spatial constraints.

AI & Developer Tools

Kyra Observability Dashboard

LangSmith-inspired observability platform for LLM applications. Track API calls, token usage, and performance metrics in real-time.

Kyra SDK (npm)

Lightweight Node.js SDK for tracing LLM calls and multi-step chains with OpenAI integration.

KitKat AI

Chat with your sticky notes. AI-powered note-taking app with real-time embeddings and contextual search.

Webhook Debugger

Lightweight webhook debugging tool that generates a unique endpoint to capture and inspect incoming HTTP requests in real time.

logsneat

Built a mini agent-observability platform on top of OpenTelemetry. A drop-in SDK (published to npm) auto-instruments OpenAI and captures your agent's full execution as a tree of typed spans — workflows, tools, retrievals, guardrails, LLM calls, which a Bun + Hono backend ingests via OTLP.

github | demo | package

GitHub PR Reviewer

AI-powered code review tool. Built to understand how automated PR reviews work under the hood.

AI Memory Layer

Mini implementation inspired by the mem0 research paper. Experimental memory system for persistent AI context.

Database & Search Tools

DataShorts→ domain fix underway

Chat with your PostgreSQL database in natural language. 4th place at HackByte 3.0 (300+ teams).

DB Chat

Converse with your PostgreSQL schema. Early version of natural language database interactions.

PlumAI

Perplexity-inspired AI search. Contextual, retrieval-augmented responses using OpenAI and Tavily Search API.

Freelance & Client Projects