Rahul Bonala
Book a sessionResume

Sri Sai Rahul BonalaPortfolio

Productdesigner forthe AI era.

I design AI-native products and write the code that ships them, because handoff is where most good ideas quietly die. Right now that's at Ginthi.ai, and on my own products like BestAnswers.AI.

Selected workGet in touch
Now
Product Designer, AI specialist at Ginthi.ai
Built
BestAnswers.AI and Stepforge, solo
Teaching
Zero to Live, 1:1 sessions for new builders
Based
Bangalore, India. Working globally.
001 About

I design the thing,
then I build the thing.

I started as a designer at Smiths Detection, drawing the service consoles that maintenance engineers across Europe, APAC, and North America use every day. Then I noticed where quality leaks out: the distance between a design and the shipped thing. So I learned to close it myself: first the React in front, then the Java and Python behind it.

That habit compounded. Today I'm the product designer and AI specialist at Ginthi.ai, designing and developing a live platform with production traffic flowing the whole time. Nights and weekends, I build with AI: BestAnswers.AI puts four models in a room and makes them argue before you get an answer.

And because the fastest way to sharpen a workflow is to teach it, I run Zero to Live, one-on-one sessions where someone with an idea and no code leaves with a website they built themselves.

Sri Sai Rahul Bonala, portrait
Bangalore, ISTRemote anywhere
95%CSAT across three regions80%faster maintenance workflows70%fewer support tickets
  1. Now

    Product Designer, AI specialist (UI/UX)

    Ginthi.ai

    I own design and front-end build for Ginthi’s web platform, taking features from concept through interface to shipped React, and building the AI tooling that lets a small team move quickly. Designing and developing the same surface means there’s no handoff to lose intent in.

  2. 2023 – 2026

    UI/UX Designer & Developer

    Smiths Detection

    Designed and built service consoles used daily by engineers across Europe, APAC, and North America. React, Java, and Python behind the screens.

  3. 2019 – 2023

    B.Tech, Electronics & Instrumentation

    SASTRA University

    Robotics and IoT on Raspberry Pi, OpenCV, and embedded C. Systems thinking that still shapes how I build.

002 Selected work

Four things
worth your time.

Build Your Own Company configurator

AI Agents · Experiment

BYOC: SDLC Agents

What if the agency was software?

Build Your Own Company turns the software delivery lifecycle into a product. Configure what your product needs brick by brick (auth, payments, dashboards) and watch the cost update in real time, then follow the build through a metro-map of SDLC phases. Underneath, AI agents carry the work through those phases autonomously.

  • AI agents mapped to SDLC phases: plan, design, build, test, ship
  • Interactive 3D-style configurator and a metro-map pipeline tracker
  • Honesty about cost as the core product idea, with no black-box quotes
ReactTypeScriptAI agentsNode
Visit live site
BestAnswers.AI interface

AI Product · Live

BestAnswers.AI

Four AIs argue. You get the strongest answer.

Ask one model and you get one confident voice, right or wrong. BestAnswers.AI convenes four AI personas (Researcher, Engineer, Friend, Academic), lets them debate in parallel, and a meta-judge merges the strongest reasoning into a single answer with the disagreements left visible. Trust through transparent disagreement, not single-source confidence.

  • Multi-agent debate architecture: parallel calls across Gemini, Llama, and Mistral
  • Meta-judge consensus layer with a transparent verdict on why an answer won
  • Built solo, end to end: product strategy, UI, and architecture
ReactTypeScriptViteTailwindMulti-LLM
Read the case studyVisit live site
Stepforge logo

SaaS · Live

Stepforge

Upload a résumé. Get a portfolio website.

Drop in a CV (PDF, Word, Pages, or a photo of a printed page) and a few seconds later you have a real portfolio site: hero, work history as case studies, projects, skills, contact. Pick from 33 designs, then download it as one portable HTML file or publish it to a shareable link. The parser is deliberately deterministic rather than a model, because a regex cannot invent an employer; AI is opt-in, is forbidden from adding facts, and never touches names, dates or job titles.

  • Extraction and parsing run in your own browser, so the résumé isn’t uploaded unless you choose a cloud feature
  • ~89% text and number retention across a 46-résumé test corpus, with a validation gate that drops anything it can’t type rather than guessing
  • Ships with its own safety nets: accessibility, responsive, print and old-mobile-engine gates in CI
ReactTypeScriptSupabaseVercel FunctionsGemini
Visit live site
your ideafree ai tools1 hourlive · togethera live website

Teaching · Live Sessions

Zero to Live

Your idea, live on the internet, in an hour.

Once or twice a week I block an hour to sit with one person, usually a student or someone early in their career, and take their idea all the way to a live website using free AI tools. Define, design, develop, deploy, in one sitting. They leave with the Playbook, every prompt we used, and the thing most tutorials never teach: how to fix it themselves when it breaks.

  • You bring an idea; we take it all the way to a live URL in the hour
  • The whole path in one go: define, design, develop, deploy, all with free AI tools
  • You leave with the Playbook and the prompts, so the next one you build alone
1:1 · liveFree AI toolsIdea → live URL
How a session works

From the archive

003 Receipts

Claims,
with proof.

No percentages, no progress bars. Every line links to the shipped thing that backs it.

Design

  • Product & UI/UX designTook a multi-agent AI product from blank canvas to shipped interface, solo.BestAnswers.AI case study
  • Design systemsService consoles built on a shared component system used across three regions.Smiths Detection case study
  • AccessibilityThis site: keyboard-navigable, AA contrast, reduced-motion honoured.Read the source

Engineering

AI

  • Multi-agent systemsFour LLM personas debating in parallel, merged by a meta-judge.BestAnswers.AI (live)
  • Agentic workflowsAI agents mapped to SDLC phases in an interactive build configurator.Build Your Own Company (live)
  • Deterministic-first pipelinesHeuristic parsing does the work; AI is opt-in and cannot invent a fact.Stepforge (live)
  • Teaching the workflowLive 1:1 sessions taking someone’s idea to a deployed site with free AI tools.Zero to Live sessions
004 Approach

Design and code are one job.

Most teams split the work: a designer hands a mockup to an engineer, and the gap between them is where intent leaks out. The spacing drifts, the empty state never gets built, the animation that made the idea legible gets dropped for time. I close that gap by doing both ends myself.

BestAnswers.AI is the clearest example. The core idea (four AI personas arguing, then a judge merging the strongest reasoning) only works if you can see the disagreement. I couldn't spec that in a static frame, so I designed it in code: built the debate view, watched four real model responses land, found that simultaneous streaming read as chaos, and redesigned the sequencing right there in the same file. Design decision and implementation were the same act. The verdict graph in this page's hero is that same idea, miniaturised.

Design a little, build a little, let the running thing correct the next decision. That loop is how everything here gets made.

005 Contact

Let's build
something real.

A role, a freelance project, an AI idea, or a question about the course. I read every email and reply within a day.

rahulbonala06@gmail.com
Book a session