hello!! i’m aimen a design engineer building atthe intersection of creativity,storytelling & tech studying cs + creative writing@ stanford

work & projects

Amazon · Summer 2026

Agentic tooling for large-scale operational data

Overview

I built agentic tooling that collapses a manual, multi-source investigation into a single query. Reconstructing why an automated decision engine produced a given outcome previously meant correlating raw logs against several disjoint datasets by hand. The tool resolves that lineage programmatically, joining the decision record to its downstream effects and returning a deterministic, evidence-backed explanation rather than an inference.

The system is structured as a transport-agnostic core with thin adapters, so query, reasoning and serialisation logic stay in one place and every surface is guaranteed to return identical results. The core exposes a backend-selectable read path over a shared dataset, and access is partitioned into tiers enforced at the code boundary, isolating privileged reads and unbounded scan costs from general traffic. Reasoning itself runs as a closed feedback loop: assemble evidence, commit to a verdict, reconcile it against observed downstream state, and fold each discrepancy back into the next evaluation.

Designing for consumers outside my own team shifted the engineering constraints from correctness to contract stability, observability and graceful degradation when an upstream dataset is incomplete. A tool that passes its tests and a tool another team depends on carry very different burdens of proof.

RoleSoftware Development Engineer Intern
TimelineJune to September 2026
Tools and servicesAWS: S3, Athena, Redshift, DynamoDB, Glue. Java and SQL over streaming and large-scale datasets. Agentic workflows via the Model Context Protocol.

Amazon · Summer 2025

Turning a support queue into self-service diagnostics

Overview

I designed and shipped a full-stack diagnostics platform that exposes real-time purchase order ineligibility reasons to more than ten thousand global vendors. Previously the reasoning behind an ineligible order was only recoverable internally, so every question became a support ticket and every ticket became an engineer manually tracing state. Surfacing that determination at its source converted a human-mediated support path into self-service, reducing redundant tickets by 76 percent.

The system is event-driven and serverless end to end: a Java service layer behind a managed API gateway, compute on demand, and a key-value store with access patterns designed around the read shapes the interface actually needed rather than the shape of the upstream data. The client is a typed React front end that renders each ineligibility reason as a discrete, explainable state instead of a generic failure. It was released through a zero-downtime deployment against live vendor traffic.

I also built an on-call automation tool outside the scope of my assigned project, applying anomaly detection over error logs to classify incidents and route them to the correct owner as they emerged. It reduced mean time to resolution by 40 percent and was subsequently adopted by teams beyond my own, which taught me that internal tooling earns adoption on clarity and reliability rather than novelty.

RoleSoftware Development Engineer Intern
TimelineJune to September 2025
Tools and servicesJava, React and TypeScript. AWS: Lambda, DynamoDB, API Gateway. Anomaly detection over log data for incident classification.

Nudgi · Behaviour design and mobile application

A gentler way to answer the message you keep meaning to answer

A messaging app built around social accountability, where a friend can send a light nudge on a conversation you have not replied to yet.

Final Nudgi screens covering onboarding, avatar selection, contacts and notification setup
The final onboarding flow: create an account, pick a penguin, find friends, then choose how you want to be nudged.

The problem

Replying to a message takes under a minute, yet these are precisely the tasks people postpone for days. My working question was why a task that small becomes that heavy. The literature pointed at the five-minute rule and at context switching, but the interesting part was not the delay itself. It was the residue: the low, continuous mental load of carrying an unanswered message around, and the guilt that grows the longer the reply is owed.

Competitive map plotting reminder apps by task context and nudge frequency
Mapping the landscape across two axes that mattered: personal against professional task context, and how frequently a product is willing to nudge. The lower-left quadrant, personal tasks with almost no nudging, was empty.

Baseline research

I ran a three-part study: a pre-study interview, a five-day diary study in which participants logged social micro-responsibilities they had delayed, and a post-study interview. The diary format mattered because these tasks are frequent and highly context-dependent, so they are almost impossible to recall accurately after the fact. One finding arrived for free: several participants reported that keeping the diary made them more responsive on its own, which was early evidence that mere awareness is itself an intervention.

Affinity clusters built from diary study and interview data
Journey map of user tasks, details and features held for later
Synthesis, then a journey map used to locate the moment of intervention. The richest opportunities sat at transition points, when someone finishes one task and decides whether to start another.

From assumption to design

Before building anything I mapped every assumption the concept rested on against two axes, how important it was and how much evidence actually supported it, then tested the ones sitting in the dangerous quadrant: important and unevidenced. The central assumption was that a nudge from a friend would read as care rather than as pressure. That single question determined the entire tone of the product.

Assumption map plotting importance against available evidence
Assumption mapping. Anything important and unevidenced was tested before it became a feature.

The resulting scope was deliberately narrow. Account creation with an avatar, an inbox with read and unread state, manual nudges sent with a single tap from inside a conversation, and automatic self-nudges scheduled for the times of day a person is most likely to reply. Nudged conversations rise to the top of the inbox, and every nudge can be answered, snoozed or switched off. The snooze was not a convenience feature; it was the mechanism that keeps the whole system from becoming another source of guilt.

Nudging flow and onboarding flow diagrams
Low fidelity screens covering receiving a nudge, disliking, responding and snoozing
Flows first, then low-fidelity screens for every branch of the nudge: received, answered, snoozed and dismissed.

Branding

The visual language had to do the emotional work. An app that reminds you of an obligation can very easily feel like a manager, so the brand leans soft and warm: a rounded, friendly type pairing, a muted palette built on dusty blue and cream rather than alert colours, and a penguin mascot that users choose and dress. Picking a penguin during onboarding is a small piece of ownership before the app ever asks anything of you.

Moodboard of soft palettes and character illustration
Nudgi style tile with colour, typography, buttons and icons
Moodboard into style tile: colour, typography, components and iconography fixed before high-fidelity design began.

Outcome

Each round of prototypes, from low to medium to high fidelity, went through usability testing, and the tone of the nudge copy was revised more often than any interface element. The finished prototype demonstrates the full loop: a conversation goes unanswered, a nudge arrives, and the reply becomes the easiest thing to do next rather than the heaviest.

RoleDesigner and researcher
TimelineWinter 2026
MethodsDiary study, affinity mapping, journey mapping, assumption testing, usability testing
ToolsFigma, FigJam

Roots · Mobile application

Helping immigrant families talk across two cultures

A mobile app of guided lessons, conversation games and a shared scrapbook, built so that immigrant parents can learn culturally specific communication skills and practise them with their children.

Roots learning screen showing Lesson 1, Understanding Conflict
Lesson detail screen offering reading, video and reflection
The learning path. Each lesson opens into a reading, a short video and a personal reflection.

Overview

Immigrant parents and their children often struggle to connect across cultural and generational distance. The premise of Roots is that this is a learnable skill rather than a fixed condition: the app teaches a communication concept, gives the family a low-pressure way to practise it together, and keeps a record of what came out of it.

The research began with nine interviews spanning parents, teenagers and a clinical expert, recruited deliberately across ages, cultural backgrounds and socioeconomic contexts. Synthesising those conversations through empathy mapping produced a consistent tension: everyone wanted closeness, and nobody had culturally specific tools for reaching it. Parents wanted to connect but did not know how to begin the conversation; children wanted the same and read the silence as disinterest.

Affinity mapping session with clustered sticky notes
Synthesis. Interview transcripts were broken into what people said, did, thought and felt, then clustered to surface the patterns.

The pivot

The most consequential decision came out of low-fidelity testing. The original concept served parent and child as equal users in one interface. Testing showed that their needs diverge sharply, and that a dual-user design actually weakened the product: parents engaged less readily when the experience was also addressed to their children, and the shared surface made the vulnerable parts feel observed rather than safe.

So the design narrowed to a single primary user, the parent, learning culturally relevant communication skills and then initiating practice with their child. Narrowing the audience made every subsequent decision easier, and it is the change that turned a broad idea into a usable product.

Low fidelity wireframes mapping the flow through onboarding, learning and games
Low-fidelity flows used to test the structure before any interface was built.

The build

I worked as an app developer on the prototype, built in React Native with Expo so that one codebase could run on both platforms without configuring native projects, and backed by Supabase for authentication, translated game content and lightweight run metadata. Internationalisation runs through react-i18next, which made a genuine language toggle possible rather than a mocked one.

The prototype implements one deliberate end-to-end path, lesson, then game, then scrapbook, in order to test the interaction model rather than to stand up a full curriculum engine. Game content is stored remotely and randomised at fetch time so that repeat sessions differ, while the lesson structure remains fixed. Deciding what to build honestly and what to stage was itself part of the work: an onboarding survey signals how the app could adapt to household context, and the documentation states plainly which behaviours are real and which are still curated.

Challenges screen, conversation starter game and a question card
Practice. Challenges unlock in sequence, and the chatting game deals question and action cards for a parent and child to work through together.
RoleApp developer
TimelineAutumn 2025
StackReact Native, Expo, Supabase, react-i18next
ResearchNine interviews, empathy mapping, low and medium fidelity testing

Dosti Project · Stanford Indo-Pak Dosti Forum · Brand identity and growth

Branding a friendship the region is taught not to want

The visual identity, merchandise and audience growth behind a student movement asking India and Pakistan to imagine a relationship built by people rather than states.

Dosti conference banner with the wordmark repeated inside a woven border

The brief

Dosti Project is a 501(c)(3) nonprofit I co-founded to build cultural connection between India and Pakistan, two countries whose citizens have spent more than seventy-five years as neighbours who cannot easily visit one another. Borders close, visas are refused, and for most people on either side the other country arrives only through a news cycle built for outrage. That was the market this brand had to enter: not a product with a clear benefit, but a relationship an audience had been taught to distrust.

The strategic decision that shaped every design choice that followed was this: lead with similarity, never with difference. The politics between the two countries are genuinely difficult, but the culture is not. The same language is spoken with a different script, the same food is argued over under different names, and the music, humour and streets are shared. The identity needed to read as home to someone in Lahore and someone in Delhi at once, without flattening either into the other. I rejected anything neutral or diplomatic in tone, because neutrality reads as distance, and distance was the problem the brand existed to solve.

The mark

I resolved that strategy into a single wordmark. Dosti means friendship, and the word belongs equally to Urdu and Hindi, so I built the logo to be read three ways at once: Latin letterforms carry the English, while their construction borrows the sweep of Nastaliq and the connected headline of Devanagari. It is deliberately one object rather than three translations placed side by side, because the argument of the movement is that these were never three separate things.

Dosti wordmark in gold on cream
Dosti wordmark in cream on gold
Dosti wordmark in plum on gold
The mark across its colourways. Plum and gold are shared ceremonial colours across the subcontinent, familiar from weddings and festival print on both sides of the border and belonging exclusively to neither.
Dosti mark set inside a repeating geometric border on plum
The mark set into a woven border drawn from textile geometry, used across print and environmental applications.

Growth

Alongside the identity, I led growth, taking the movement from a standing start to an audience of more than twelve thousand across both countries and the diaspora. My content strategy followed directly from the brand strategy: publish the overlap. Shared vocabulary, shared recipes, shared jokes, and split screens of two cities that look like one. I avoided political content entirely, not out of caution but because comment sections punish it while recognition travels. The best-performing posts were the ones where someone in Karachi and someone in Bombay tagged each other under the same image.

Dosti project Instagram profile showing more than twelve thousand followers
@dostiproject on Instagram.

Merchandise

I applied the identity to objects people actually keep. One tote carries an illustrated inventory of things both countries claim and share: karak chai in its glass, a kite, a lotus, a dhol, a decorated rickshaw, a cricket bat, a mango. Another prints a dictionary entry for the word itself, defining dosti as chai, butter chicken, doctor bano ge ya engineer? and aur sunao?, until it is time for cricket.

Printed Dosti tote bags photographed on a table
The finished totes.
Tote bag printed with chai, a kite, a lotus, a dhol and a rickshaw
Tote bag printed with a dictionary definition of dosti
Tote bag printed with hasso, jiyo, dosti karo, kia pata, kal ho naa ho
Tote bag printed with the wordmark repeated in four colourways
Maroon crewneck with the Dosti wordmark
Back of the crewneck with the hasso jiyo lyric and block-print sleeve detail
The wider range. Nothing in it names a state; everything in it names a childhood.

Conference

The identity now carries the inaugural Stanford Indo-Pak Dosti Conference at the Hoover Institution, convening former heads of government, founders, investors, filmmakers and artists around people-led partnership. I extended the same system across the website, the printed programme and the physical environment.

The live conference website showing the speakers section
The site as built, at indopakdosti.org.

I designed and set aside two earlier directions. The first leaned into plum as a full-bleed field, which read as formal and institutional, closer to a summit than a movement. The second framed a light page with the border motif running down both edges, which was warmer but pulled attention towards the ornament rather than the speakers. The published site keeps the motif as an accent and gives the room to the people in it.

Unused website direction on a full plum background
Unused website direction, light page with motif borders down each side
Directions not taken.
OrganizationDosti Project, a 501(c)(3) nonprofit
RoleCo-founder. Design lead and growth lead
TimelineNovember 2023 to present
ScopeVisual identity, wordmark, merchandise, social strategy, conference collateral
Reach12,000+ following. $50,000 raised. Two student-initiated courses taught to more than ninety students

explore my creative playground

coming soon

experiments, side projects and a cat sticker for every emotion. currently on the loom.

about

a cross-stitch is a grid of small x’s, counted and placed one at a time. a screen is a grid of small squares, switched on one at a time. the same instinct runs through both: decide what belongs in each cell, and an image appears. i grew up around phulkari, the counted-thread embroidery of Punjab, where every stitch is placed by counting the threads of the fabric itself. i later studied computer science, and it took me a long time to realise the two were the same craft.

that idea shapes what i build. i am a design engineer studying computer science and creative writing at Stanford, and i am drawn to work where the engineering is precise and the feeling is the purpose. i like interfaces that carry a story, and stories that need some engineering to be told well.

the thread running through all of it is storytelling, in whichever medium suits the story: illustration, film, writing, or code. what i am really chasing in each of them is human emotion, the small honest moments people usually keep to themselves. i build creative technology that makes room for that feeling rather than designing it away, and above all i want the things i make to feel as though a person made them.

away from the screen, i am a devoted chai drinker: doodh patti, brewed longer than it should be, ideally in good company. i skate, i make pottery, and i am happiest doing anything that leaves my hands messy. i am also, reliably, the friend with a cat sticker for every possible emotion, which is either a personality trait or a filing system. see the cat sticker collection →

aimen, rendered as a dithered bitmap portrait
one motif, two crafts. drag to cross the border between thread and pixel.