Shortlist01
Margin02
FITFLOW AI03
Creators Store04
COMET 2.005
Solune & Afterdark06Selected work · 2025–26
Six products, one habit: show the evidence.
Product Designer · AI products · Bengaluru
Three things, in one person
01
Product, UX, systems, motion
From the problem and the user flow to the interface, the design system and the motion: dark and light themes, Figma variables mirrored in code, and interactions that explain themselves.
02
LLMs, prompting, evaluation, RAG
A year evaluating LLM outputs for accuracy, factual consistency and safety, then leading a team on a client project for Amazon AGI Lab. In my products: hybrid retrieval, structured outputs, quote verification and labelled evaluation sets.
03
Next.js, TypeScript, Python, FastAPI
Every project here is live. Full-stack products with tests and CI, payments verified on the server, WebGL and path tracing in the browser. I build with AI coding agents, and I review the code and the evaluations.
Selected work · AI products
Each one is live. Open a case study for the problem, the decisions, the architecture, the AI and the evidence.
Lab · brand, motion and WebGL
Approach
01 / 05
Every AI claim links to the passage it came from, and the quote is checked against the source before it's shown.
Seen inMarginShortlist
02 / 05
Users can see what the document says and what the model inferred, labelled on every claim.
Seen inMargin
03 / 05
When the sources don't support an answer, the product says so instead of guessing, and shows what it searched.
Seen inMarginShortlist
04 / 05
Nothing is sent, published or overwritten without a person choosing to. Drafts explain their reasoning.
Seen inFITFLOWCreators Store
05 / 05
I build labelled test sets and report the results, including the failures and the caveats.
Seen inShortlistMargin
Experience
Open any chapter. The AI years at Keywords Studios are the reason my products show their evidence.
Machine learning and software engineering, and two published papers before graduating.
Eight months of full-stack training, then an internship building a responsive storefront.
Generative AI model training: writing prompts, evaluating responses, annotating data and keeping datasets clean.
01
Model training & response evaluation
02
Dataset quality
03
Documentation
Promoted after four months. Led a team of Research Associates on a client project for Amazon AGI Lab.
01
Prompt engineering & optimisation
02
LLM output evaluation
03
Hallucination & edge-case analysis
04
Training data & team lead
Operational and revenue analytics for the software behind visitor services at Global Village UAE.
Bengaluru, India
About
At Keywords Studios I reviewed model outputs for accuracy and safety and traced the patterns behind hallucinations, then led a small team doing that work on a client project for Amazon AGI Lab. It taught me that the hard part of an AI product isn't the model; it's helping people know when to trust it.
So I design AI features around what models get wrong: evidence you can open, fact kept apart from judgement, an honest “I don't know”, and a person who makes the final call. I take an idea from the problem and the user flow to the interface, the design system and a working, deployed product.
I build with AI coding agents, mainly Claude Code. I choose the problem, set the product direction and the design, review the code and the evaluations, and decide what ships. Ask me about any decision in this portfolio and I'll walk you through it.
Contact
I'm looking for Product Designer and AI Product Designer roles, where I can own an AI feature from the first sketch to what ships. The quickest way to reach me: