---
title: "About Harshith Nayaka L — Background, Stack and Published Research"
description: "Who Harshith Nayaka L is: his role at DemandNXT, what he has built since 2022, his stack, his co-authored IJRTMR paper, and answers to common questions."
canonical: "https://harshith-nayaka-l-portfolio.vercel.app/about"
last-updated: "2026-10-31T06:08:19Z"
author: "Harshith Nayaka L"
author-role: "AI Engineer"
author-title: "AI Engineer"
author-employer: "DemandNXT"
author-location: "Bengaluru, India"
author-availability: "Available for freelance work"
author-email: "harshith28124@gmail.com"
content-type: "text/markdown"
html-version: "https://harshith-nayaka-l-portfolio.vercel.app/about"
---
# About — Harshith Nayaka L

> Who Harshith Nayaka L is: his role at DemandNXT, what he has built since 2022, his stack, his co-authored IJRTMR paper, and answers to common questions.

Canonical page: https://harshith-nayaka-l-portfolio.vercel.app/about

Harshith Nayaka L is an AI Engineer in Bengaluru (Bangalore), India. He leads AI engineering at DemandNXT, where he owns AI work across the company's six B2B brands. He builds AI agents, RAG pipelines and AI workflow automation for marketing and creative operations, and is available for freelance work.

## At a glance

AI Engineer at DemandNXT · Bengaluru (Bangalore), India · Available for freelance work.
Specialties: AI agents and multi-agent systems, RAG pipelines, AI workflow automation with n8n, and LLM apps built end to end.
Shipped: Octo, a live multi-agent research workspace on the OpenAI Agents API; SPECTRA, a live AI search visibility audit; Personal MCP OS, an 85-tool local MCP server where risky actions wait for one-time human approval; Maestro, a live open-source multi-model LLM orchestration engine; and a roughly 90-node n8n content pipeline with QA gates.
Across these projects, Harshith Nayaka L's stack is Python and FastAPI; TypeScript, React, Next.js and Node.js; n8n for workflow automation; Google Gemini, Groq-hosted open models and OpenAI models; PostgreSQL with pgvector, SQLite and Redis; the Model Context Protocol; and Playwright and Vitest for testing.

## The short version

Role: AI Engineer at DemandNXT. Based in Bengaluru, India. Building since 2022. Available for freelance work.
The work is AI agents, retrieval pipelines and AI applications built end to end — the model, the backend, and the interface around them. At DemandNXT he leads AI engineering: production AI systems and pipelines for marketing and creative operations across its six B2B brands. He is also migrating DemandNXT's own website from legacy WordPress to a codebase built with Astro.

## Selected work, one line each

Octo (Multi-agent research workspace): Multi-agent research on the live web with cited reports it checks itself, and a Gemini Live voice assistant that sets up research but never starts it.
SPECTRA (AI search visibility audit): Asks AI the questions your buyers ask and shows whether your site is in the answer — then traces what a model actually understood about the site, and at which stage a fact disappeared. The model emits typed observations; a deterministic engine does the scoring.
Personal MCP OS (85-tool local MCP server): Gives an MCP client real hands on a machine — filesystem, shell, Git, browser, Android — behind risk-based policy and one-use approvals a human has to grant.
BrandForge (AI brand campaign pipeline): Give it a brand's site and one product photo: it researches the brand, pins what can't change about the product, generates a six-shot campaign, then grades its own output and blocks what it can't safely fix.
Maestro (Multi-model LLM orchestration): A glass-box engine that routes one task across free LLMs — conductor, thinker, worker, verifier — and returns a single verified answer with a full decision-log of every step.
Multi-Brand Audit Platform (SEO audit tooling): An internal audit platform a marketing team runs on across a multi-brand portfolio — real-browser crawling, and a score where every weight traces to something Google actually published.
Creative-Ops Pipeline (~90-node n8n workflow): A multi-model content pipeline that turns a one-line brief into validated, on-brand output — tiered model routing, schema-constrained generation and QA gates, orchestrated across roughly ninety n8n nodes.
CraftConnect (Multi-modal artisan assistant): A multi-modal assistant that lets artisans run an online storefront by talking and showing, not typing.
Nova (Cost-tiered model routing): A chat app that scores every turn for difficulty and sends it to the smallest model that can carry it — lane, reasoning and cost stamped on every answer.
Cannon (Multi-agent personal assistant): A daily-driver multi-agent assistant where each domain gets its own specialist — own persona, own tools, own retrieval scope, isolation enforced at the query, not by convention.
AI Company Briefs (n8n workflow: Airtable to Google Docs): Every morning, a ChatGPT brief on each company in an Airtable contact list, saved as its own Google Doc, with a Slack digest that links them all and names any company it couldn't finish.
YouTube to Sheets Scraper (n8n YouTube Data API scraper): Search terms, videos, playlists and channels listed in a Google Sheet become one de-duplicated row per video, through the official YouTube Data API rather than a scraper that breaks.
Slack Weekly Report Bot (n8n weekly outreach report): Posts the week's outreach numbers to a private Slack channel every Monday, and refuses to post stale weeks, half-filled rows or to a public channel.
ReplyDesk (WhatsApp lead agent): An interactive prototype of a WhatsApp lead agent: simulate an inbound lead and watch it get a first reply in ~8 seconds, with a live dashboard for the lead feed, pipeline, and response-time KPIs.
AI Notes (Local-LLM notes app): A notes app running a local LLM through Ollama, so inference never leaves the machine.

## What the work actually is

Most of the engineering in an AI system is not the model. It is everything that assumes the model can be wrong: routing a request before spending on it, constraining output to a schema and validating it anyway, having a second model on a different family verify the first, isolating agents so context cannot leak between them, and keeping a decision log that can be replayed after something goes wrong.
That is the difference between a demo that works once and a system a business can lean on, and it is most of the job. There are 15 case studies on this site and each one is a version of that argument with the code attached.

## Everything here is checkable

Building with code since 2022: a few years of shipping things, breaking them and learning what reliable looks like up close.
Everything stated here is checkable. Most projects are public repositories (the two internal ones, the Multi-Brand Audit Platform and the Creative-Ops Pipeline, say so on their pages), the research is a published paper with a DOI, and where a number appears it traces to something that can be read rather than taken on trust.

## Experience

AI Engineer - Full Stack, DemandNXT (current role). Leads AI engineering across DemandNXT's six B2B brands.
AI Workflow Engineer, DemandNXT (previous role). Moved from this role to AI Engineer.

## Published research

Co-author of “AI-Powered Note-Taking System: A Local Machine Learning Approach DeepSeek R1 Integration”, published in the International Journal of Research Trends and Multidisciplinary Research (IJRTMR), Nov–Dec 2025. It documents the local-inference approach behind the AI Notes project: summarisation running on DeepSeek R1 through Ollama, with no cloud fallback.

## Questions people ask about Harshith Nayaka L

### Who is Harshith Nayaka L?

Harshith Nayaka L is an AI Engineer in Bengaluru (Bangalore), India, where he leads AI engineering at DemandNXT. He builds AI agents, RAG pipelines and AI workflow automation, along with the backends and interfaces around them, and he is available for freelance work.

### What does Harshith Nayaka L do?

Harshith Nayaka L builds AI systems that have to hold up in production: AI agents and multi-agent systems, RAG pipelines, AI workflow automation with n8n, and LLM apps built end to end, backend and interface included. At DemandNXT he builds AI systems and pipelines for marketing and creative operations.

### What is Harshith Nayaka L's current job title?

Harshith Nayaka L is an AI Engineer at DemandNXT in Bengaluru; in practice he leads the company's AI engineering, across DemandNXT's six B2B brands. He moved into this role from his previous role at DemandNXT, AI Workflow Engineer.

### Where is Harshith Nayaka L based?

Harshith Nayaka L is based in Bengaluru (Bangalore), Karnataka, India.

### What projects has Harshith Nayaka L built?

Harshith Nayaka L has 15 projects with published case studies: Octo (Multi-agent research workspace), SPECTRA (AI search visibility audit), Personal MCP OS (85-tool local MCP server), BrandForge (AI brand campaign pipeline), Maestro (Multi-model LLM orchestration), Multi-Brand Audit Platform (SEO audit tooling), Creative-Ops Pipeline (~90-node n8n workflow), CraftConnect (Multi-modal artisan assistant), Nova (Cost-tiered model routing), Cannon (Multi-agent personal assistant), AI Company Briefs (n8n workflow: Airtable to Google Docs), YouTube to Sheets Scraper (n8n YouTube Data API scraper), Slack Weekly Report Bot (n8n weekly outreach report), ReplyDesk (WhatsApp lead agent), AI Notes (Local-LLM notes app). 13 of them link to public code on GitHub (https://github.com/HarshithNayakaL), and every case study is on his site.

### Does Harshith Nayaka L build AI agents and RAG pipelines?

Yes. Harshith Nayaka L builds AI agents, multi-agent systems and RAG pipelines. Octo is a live research workspace where a director agent delegates to specialist agents on the OpenAI Agents API; Maestro routes one task across thinker, worker and verifier models; Personal MCP OS is an 85-tool MCP server where risky actions wait for human approval; and Cannon is a multi-agent assistant where each agent has its own tools and its own retrieval over Postgres with pgvector.

### Does Harshith Nayaka L work with n8n?

Yes. Harshith Nayaka L builds AI workflow automation in n8n. His Creative-Ops Pipeline is a roughly 90-node n8n workflow with model routing, schema-constrained output and QA gates, and BrandForge's orchestration can run on n8n. His public n8n workflows include AI Company Briefs (Airtable contacts to a ChatGPT brief per company, saved as Google Docs with a Slack digest), a YouTube to Google Sheets scraper on the official YouTube Data API, and a Slack weekly report bot that refuses to post stale or incomplete numbers, each tested live on n8n 2.8.4.

### What is Harshith Nayaka L's tech stack?

Harshith Nayaka L works in Python and FastAPI; TypeScript, React, Next.js and Node.js; n8n for workflow automation; Google Gemini, Groq-hosted open models and OpenAI models; PostgreSQL with pgvector, SQLite and Redis; the Model Context Protocol; and Playwright and Vitest for testing.

### How much experience does Harshith Nayaka L have?

Harshith Nayaka L has been building software since 2022. He is an AI Engineer at DemandNXT, where he leads AI engineering; he has 15 projects with published case studies, and co-authored a peer-reviewed paper on local LLM inference (IJRTMR, 2025).

### Is Harshith Nayaka L available for freelance work?

Yes. Harshith Nayaka L is available for freelance work alongside his role at DemandNXT. The direct route is email: harshith28124@gmail.com.

### How can I contact Harshith Nayaka L?

Email Harshith Nayaka L at harshith28124@gmail.com, or use the contact form on his site. His GitHub is https://github.com/HarshithNayakaL and his LinkedIn is https://www.linkedin.com/in/harshithnayakal.

### Has Harshith Nayaka L published any research?

Yes. Harshith Nayaka L co-authored “AI-Powered Note-Taking System: A Local Machine Learning Approach DeepSeek R1 Integration”, published in IJRTMR in Nov–Dec 2025 (DOI 10.59256/ijrtmr.20250506023). It documents the approach behind his AI Notes app, which runs DeepSeek R1 locally through Ollama.