GayatriSrinivasan^{Frontend developer & designer}

Most of what I build sits on top of a model. I design the layer between what a system predicts and the person who has to act on it.

1. 01
   Plate 01Personal finance tracker

   Nook/

   A mobile-first finance app where the AI never touches the arithmetic. Every number is computed in SQL; the model only ever sees pre-computed facts.

   Built with

   Next.js · React · Supabase · Tailwind · Design tokens · Groq

   Live Code

   Read case study
2. 02
   Plate 02Two-player colour guessing game

   Hue Knew?/

   A colour guessing game for two, inspired by Hues and Cues. Real-time multiplayer over WebRTC, phones as private hands while a laptop plays the table, and every sound synthesized in code.

   Built with

   Vanilla JS · WebRTC · PeerJS · Web Audio · OKLCH

   Play Code

   Read case study
3. 03
   Plate 03iOS widget

   Daysie/

   An event countdown widget with 1,200+ valid configurations across background, countdown type, colour, font, symbol and size, constrained so no combination produces an unreadable result.

   Built with

   Swift · Constraint design

4. 04
   Plate 04Inception World Model Hackathon II

   Different Worlds/

   A game where each level is generated by a world model and players progress by finding objects that share a hidden theme. Built in a weekend as a first attempt at world models, and featured afterwards in the Reactor newsletter.

   Built with

   Reactor · Alakazam Studio · Rapid prototyping

5. 05
   Plate 05Credit scoring for thin-file users

   PsyCred/

   A credit assessment for people with no borrowing history, pairing an LLM-driven psychometric test with a demographic model. I built the frontend and designed the test experience — the interesting problem was making an assessment that decides your creditworthiness feel neither invasive nor like a trick. Built at a hackathon with the CreditIQ Crew.

   Built with

   Frontend · UI/UX

   Code
6. 06
   Plate 06Conversational support bot

   Zenon/

   A Telegram bot using prompt-chained LLMs to respond to emotional context and surface relevant resources. A student exploration of conversational design and prompt chaining rather than a product.

   Built with

   Python · Conversational design · Prompt chaining

1. Jun 2025 — Present01

   PayIntelli/

   Founding member — Frontend Developer & UI/UX Designer, AI Solutions Delivery

   • Built PayIntelli's brand identity from scratch: logo, brand guidelines and a visual system used across product, web and team communications.
   • Designed UI systems and 50+ reusable components used across 15+ product modules.
   • Developed and launched 4 customer-facing landing pages in TypeScript and Next.js.
   • Ran 15+ project workflows in Notion and ClickUp across a 25+ person cross-functional team spanning engineering, design and AI.

   Read the PayIntelli case study
2. Feb 2025 — May 202502

   PXP Financial/

   Business Intelligence Intern

   • Built a React dashboard to configure an ML fraud-detection model and read KPIs across 500k+ transactions at sub-50ms latency.
   • Trained a model that raised recall 20% over the legacy system while holding false positives under 1%.
   • Designed 4+ dashboard pages surfacing flagged transactions, model performance metrics, and weekly and monthly summaries.

   Read the PXP Protect case study
3. Mar 2024 — May 202403

   InLustro/

   Full Stack Developer

   • Built 10+ reusable React components, cutting development time roughly 30%.
   • Designed the login, signup and dashboard interfaces carrying the core user flow.
   • Built the Syntra test interface; navigation changes reduced drop-off by 20%.
   • Integrated PostgreSQL and improved backend efficiency and maintainability.

   Read the Syntra case study
4. Aug 2023 — Mar 202404

   InLustro/

   Frontend Developer

   • Built the React interface for a Google Meet plugin: login, signup, goal progress and meeting summaries, translated from Figma into responsive production UI.
   • Designed integration concepts for Slack and Microsoft Teams.

Starting point

Most of my work starts with a screen people do not trust — a dashboard with too many numbers, an internal tool nobody wants to open, a flow that loses users halfway through.

Approach

I shorten the distance between a question and a usable answer. Fewer elements, a component system the whole team can reuse, and front-end code that holds up against real data and real devices. I design in Figma and ship in React, which means I do not hand over designs that cannot be built, and I do not build interfaces I cannot defend.

Why the AI part matters

My degree is in artificial intelligence and machine learning, and most of the interfaces I have built sit on top of a model. That turns out to matter more than I expected: I can read what a model is actually telling me and design for uncertainty, rather than flattening a confidence score into a yes or no.

Most of what I have built sits on top of a model — fraud detection at PXP, a spending coach in Nook, credit scoring for people with no credit file, traffic violation detection I published and patented at university. The interface layer is where a prediction becomes a decision someone has to live with, and that is the part I find most interesting to design.

Education

B.Tech, Artificial Intelligence and Machine Learning · MLR Institute of Technology, Hyderabad · May 2025

Certifications

• Certified Associate in Project Management (CAPM) — coursework, PMI, 2026
• Start the UX Design Process: Empathize, Define, and Ideate — Google, 2024
• Foundations of User Experience (UX) Design — Google, 2023

Tell me what you are building and who has to use it. I reply to everything within two working days.