Unmarketed Labs
Case study · 2026

TechShorts

BrandPlatformAIFor IntelliFunnel Labs

From no brand to production in four weeks: an AI news feed that pulls four sources, removes the duplicates and summarises every story in exactly 60 words.

BrandDesign systemFull-stackAI pipelineAdmin dashboard
Top stories: every card is a 60-word AI summary with its source.techshorts.io

An AI-powered news platform. No human editorial step.

Rishabh Shetty, founder of IntelliFunnel Labs, needed more than a website. He needed a complete brand identity and a production-ready platform: one that could aggregate tech news from multiple premium sources, generate concise AI summaries for each article, and present them through a clean, mobile-first card interface, with no manual editorial work in between.

The platform also needed a whitepaper lead-capture system and a full admin dashboard for content management and analytics. The target: brand, design system and production platform, all in four weeks.

What it had to do
  1. 01

    Aggregate tech news from multiple premium sources.

  2. 02

    Generate a concise AI summary for every article.

  3. 03

    Present it all in a clean, mobile-first card interface, with no manual editorial work in between.

  4. 04

    Capture leads with whitepapers, and run everything from a full admin dashboard.

  5. 05

    Ship the brand, the design system and the production platform in four weeks.

The pipeline

How a story reaches the feed.

Every card in TechShorts passes through the same five steps. Nobody rewrites it by hand on the way.

  1. 01 · Fetch

    Four sources

    New stories are pulled every 30 minutes.

    PRNewswireNewsData.ioCurrentsGNews
  2. 02 · Deduplicate

    One story, once

    Hash-based and semantic deduplication, so the same story never appears twice, whichever source it came from.

    HashSemantic
  3. 03 · Summarise

    Exactly 60 words

    Groq runs Llama 3.1 70B to write a summary of exactly 60 words. Not approximately.

    GroqLlama 3.1 70B
  4. 04 · Cache

    Written once

    Each summary is cached permanently after its first generation, so AI cost does not scale with traffic.

    Cached permanently
  5. 05 · Publish

    One card

    The story lands in the feed as a card: the 60-word summary, with its source.

    AllAI TechFinTech
01 · Week 1

A brand from zero

No existing assets, no style guide, no direction. We built the identity system, logo, colours, type scale and component library, before writing a single line of code.

02 · Week 2

Four sources, no duplicates

PRNewswire, NewsData.io, Currents and GNews, every 30 minutes, with hash-based and semantic deduplication so the same story never appears twice.

03 · Week 2

Exactly 60 words

Groq (Llama 3.1 70B) writes summaries of exactly 60 words, not approximately, cached permanently after the first generation so AI cost does not scale with traffic.

Design system

Built before the first line of code.

The identity, the colours, the type scale and the component library, in the product's own tokens. Two themes share one red accent, which also marks the active category.

LogoDark · Light
TechShorts logo on the dark themeTechShorts
TechShorts logo on the light themeTechShorts
TypeSystem sans · Georgia
Aa.
Aa
Top StoriesSans 800A headline set in Georgia, on the desktop cardGeorgia 700Source, category and timeSans 500The summary, in the system sans at body sizeSans 400
The interface uses the reader's own system fontWeights in use: 400 to 800
ColourTwo themes, one accent
Dark
Background
#0F172A
Card
#1E293B
Border
#334155
Text
#F8FAFC
Primary
#EF4444
Light
Background
#FFFFFF
Card
#F8FAFC
Border
#E2E8F0
Text
#0F172A
Primary
#EF4444
ComponentsCategory bar · Story card

01Source, category and time above the story.

02The image, with its category pill.

03The headline, then the 60-word summary.

04Tap to read the full article.

The summaryA fixed length
60words in every summary. Not approximately.
Generated once by Llama 3.1 70B on GroqCached permanently
The TechShorts resources page on a laptop: one form opens the whole whitepaper library
Resources: the whitepaper library that captures leads.techshorts.io/resources

Delivered on time. Every feature shipped.

The platform launched on schedule at the end of week four, production-deployed and handed over with complete brand guidelines, technical documentation, a training session and 30 days of post-launch support.

The AI pipeline was processing articles from all four sources within 48 hours of the cron job going live. Alongside the consumer product we built a full role-based admin panel for article review, approval, lead management and live analytics, with CI/CD and monitoring, and it was adopted without a single support request after the training session.

Eight major outputs
I came to Utathya with a brief, a budget, and a 4-week deadline. Most agencies would have told me to be realistic. He told me what was possible and then delivered exactly that: brand, platform, and a working AI pipeline, all production-ready. I have not had to touch the codebase since launch.
Rishabh Shetty, Founder · IntelliFunnel Labs
By the numbers
4weeksFrom no brand to a production platform, launched on schedule.
4sourcesPRNewswire, NewsData.io, Currents and GNews, deduplicated into one feed.
60wordsIn every AI summary. Exactly, not approximately.
30minBetween fetches from every source.

Up next

EverwedPune first, or it would be everything and nothing.
Blown DiffuserNothing ships without a named human reading it.
CR NewsOne editor should never be the only lock.

Have something to build? We'd love to hear about it.

Start a projectBook the AI AuditBook a call