Shafeen Alam

Playground

Things I make because I am curious.

Products I ship on my own, AI interactions I prototype to understand them, machines I take apart, and drawings I have been making since long before I designed software. None of it is client work. All of it is where the habits come from.

I hold a BFA in Illustration and Animation from San Jose State University and worked as an illustrator and concept artist before I designed software. Thumbnails before rendering. Silhouettes before detail. The same subject three times under three different constraints, because the constraint tells you which parts actually matter. Every one of those is something I still do, just with interfaces now instead of paint.

Products

Things I designed and built myself, running in the world rather than described in a deck.

Live productSelf-initiatedBuilt solo

Stream Find

Streaming availability ranked by what you already pay for. Included options first, then free, rental, and purchase in one vocabulary. Designed and coded by me, with AI assistance throughout the build.

Open Stream Find Case study
In progress

What comes next

The next product will most likely sit in mobility. Vehicle health explanation, charging reliability, or the service and repair loop. When it exists as something you can use, it will show up here first.

Not yet public

AI experiments

Working interactions I build to understand how these systems should behave, especially at the edges where they are wrong, slow, or unsure.

Working interactionRecreation

The Interpres AI interaction

A working recreation of the assistant I designed: structured reasoning, evidence links, confidence, the branch where it refuses to answer, and the confirmation every sensitive action passes through. Play it through both branches.

Try the interaction Case study
Ongoing

AI-assisted design practice

Claude and Claude Code, Cursor, ChatGPT, and Figma Make, used to prototype workflow logic and pressure-test edge cases rather than to generate screens. Stream Find is the largest thing this practice has produced so far.

See how it was built

Mobility & Machines

Cars are the most demanding consumer system I interact with: connected hardware, software, diagnostics, safety states, and a service experience most owners never fully understand. I am interested in the same things here that I am interested in at work, which is how a complex machine explains itself to the person responsible for it.

Systems thinking beyond software

The same mindset I bring to product, applied to vehicle integration. I retrofitted driver-assistance hardware into a car that was never built to take it, which meant working inside someone else's architecture and proving each signal before trusting it.

01

Retrofit

Heads-Up Display plus Adaptive Cruise Control and Pre-Crash components, pulled into a 2013 Lexus GS450h that never shipped with them.

02

Constraints

Everything had to work inside the existing CAN bus wiring. No clean slate, no spare bandwidth, and no documentation written for this.

03

Dependencies

Sensor dependencies and control module integration, mapped before anything was spliced.

04

Validation

Signal paths verified before system activation, because a driver-assistance feature that activates on a bad signal is worse than one that never activates.

What it reinforced

Automation must be grounded, validated, and controlled before it earns trust.

Which is the same argument I made for an AI assistant in a security operations center, a few years and one industry away. It is why autonomy validation resonates with me.

The interior of a 2013 Lexus GS450h mid-retrofit, dashboard removed, wiring looms exposed across the footwell and seat, with tools and control modules laid out
Mid-retrofit. Every one of those looms had to be traced and validated before a single assistance feature was allowed to switch on.

Explainable diagnostics

A warning light is a notification, not an explanation. What would it take for a vehicle to state severity, confidence, and consequence in terms an owner can act on.

Serviceability

What the vehicle, the app, and the service center each know, and where the handoffs between them drop information the owner needed.

Human machine interaction

Controls, feedback, and failure states, judged the way I would judge any operational interface: can you tell what is happening, and what to do next.

Garage notes

A running visual journal of what I maintain, modify, and investigate, plus interface critiques from the driver's seat. My build thread lives on Instagram, and the written entries land here. Every entry answers the same five questions, which is the same structure I use for a product critique.

  1. What did I notice?
  2. Why did it matter?
  3. What did I investigate?
  4. What did I change or learn?
  5. How could the experience be improved?
@no.msg.added Check out my build

Art

Illustration and concept work, mostly predating my product career. Grouped by what each series was teaching me.

Architectural studies

Graphite, drawn from observation. Slow, and useful for exactly that reason.

Notre Dame Cathedral and a historic church Two approaches to the same subject. The left drawing crops in tight on the rose window and lets the tracery carry it. The right pulls back and uses the tree to frame the mass. Both are patience more than technique.

Still life under constraint

The same setup painted three times, each under a different palette restriction, with progress states kept alongside. This is the series I think about most in product work. The constraint is not a limitation to work around, it is what forces a decision about what actually matters in the image.

Still life, master palette Glass and reflective metal against near-black. With the full palette available, the problem is restraint: knowing which highlights to keep and which to let go.
Still life, Van Gogh palette The same objects under a historically restricted palette. Six progress states, because the interesting part is watching where the constraint forces a different choice than instinct would.
Still life, limited palette Nearly monochrome. Once hue is removed as a tool, everything has to be carried by value, and the composition either holds up or it does not. This one taught me more than the other two.

Observational drawing

Animals from life. No composition to hide behind, no rendering to fall back on. Just whether the structure underneath is right, drawn enough times that the shapes become known rather than copied.

Cows Six angles, one rendered. The construction lines stay visible on purpose, because they are the part worth keeping.
Deer Weight and balance, mostly. The leaping figure at the top is the only one that needed to be dynamic, and it is the loosest drawing on the page.
Elephants, structure Working out how the mass sits on the legs. The front-facing study is the hardest angle and the reason the sheet exists.
Elephants, gesture Faster and looser than the first sheet. Fewer marks, more information about how the trunk actually moves.
Ring-tailed lemur The hands were the problem. Everything else on this sheet is in service of the two studies where they are worked out.
Ring-tailed lemur, continued A second pass on the same animal. Drawing something twice is how you find out what you assumed the first time.

Character development

Three characters from Browning's My Last Duchess and two from Bradbury's The Veldt. Each sheet works the same way: silhouettes first to test whether the character reads at a glance, then structure, then one rendered pass. The silhouette row is the real work. If the shape does not read in black, no amount of rendering saves it.

Alfonso II D'Este, Duke of Ferrara The Duke had to read as controlling before he read as wealthy. The silhouettes test posture and the weight of the cape, so the rendered version only has to confirm a decision already made.
Lucrezia Di Cosimo De'Medici Fifteen silhouettes to find one gesture. The fan and the turned shoulder carry the whole character, which is why they survive into the final and most of the variations do not.
Fra Pandolf The painter in the poem, seated and absorbed. Three rendered passes at different fidelities, kept side by side so the loose version stays available as a reference for what the tight one is trying to preserve.
David McClean Expression studies first, because the psychologist's whole function is reading other people. The patterned jacket and bow tie arrived last, once the face was settled.
George Hadley Five heads to find the smile. Same features throughout, but the expression shifts from open to unsettling, which is the arc the character actually needs.

Environment concepts

Environments for A Day in the Life of Beethoven. Every sheet runs the same sequence: ten thumbnails to find the composition, four color keys to find the mood, then one resolved painting. Cheap exploration first, expensive rendering last, and only after the decision is already made.

Nature Walk, exterior The figure is almost incidental, which is the point. Ten thumbnails to establish that the trees should crowd the frame, then four keys to choose fog over autumn, dusk or green.
Nature Walk, vista Same character, same day, opposite feeling. The keys carry the argument: teal reads as cold, purple as elegiac, and the composition works either way, which is what makes it a good composition.
Heiligenstadt at night Two figures, one lamp-lit street, and the whole mood resting on how much blue sits in the shadow. The thumbnails here are perspective problems rather than composition problems.
Cathedral ruins A single figure under a shaft of light in a collapsed nave. The keys test whether the space should feel warm and consoling or cold and abandoned. Warm won.

Where it meets the product work

One artifact that belongs to both practices.

iOS gesture library Drawn as a reference set for documenting touch interactions. Same discipline as the character silhouettes, just applied to interface behavior: strip it to line, make each state distinguishable at a glance, and define the whole matrix rather than the one case you happen to need today.

Publications

Where this work has been published.

The Big Book of Contemporary Illustration

Martin Dawber. My cow and architectural studies appear in this collection.

Cover of The Big Book of Contemporary Illustration by Martin DawberSpread from The Big Book of Contemporary Illustration showing built environment illustrations

2009 Vans’ Warped Tour: Guide to Not Being a Complete Environmental Disaster

A full color mini magazine. My cow study appears in it, alongside original artwork by Matthew Gray Gubler, with contributions from tour creator Kevin Lyman, Matt Wachter, Philippe Cousteau, Robert Kennedy Jr.’s Waterkeeper Alliance, David Pirano, Alex Hallatt and Joshua Katcher, aka The Discerning Brute. It is packed with tips on green living, plus anecdotes on touring and living green from Warped Tour bands, past and present.

Vans Warped Tour 2009 logo

Extreme Metaphors: A Launchpad for Other Explorations

Simon Sellars. My illustration for J.G. Ballard's short story "The Drowned Giant" appears in his introduction to Extreme Metaphors: Interviews with J.G. Ballard 1967 to 2008, as published on his site.

Read the article →
Grayscale illustration for The Drowned Giant: a man with a briefcase walks toward a giant's body washed up on a beach

Interview exercises

Design assessments I was asked to complete as part of hiring processes. Short briefs, tight time, and no access to the people who built the product.

April 2008Design assessment

eBay Motors search

The brief: improve how people search for a used car on eBay Motors. I reviewed Autotrader and Cars.com, sketched ideas freehand, and tested the current site with someone new to eBay. They were overwhelmed by the number of results and missed the "More Choices" refinements. My answer was to group results into visual steps, first by model year, then mileage, then color, with each choice kept as a removable keyword above the results.

Process sketches and four-step approach1 · ProcessProblem assessment of the eBay Motors search results page2 · Problem assessmentSearch results grouped by model year3 · Grouped by yearSearch results grouped by mileage4 · Grouped by mileageSearch results grouped by color5 · Grouped by color
GivebutterDesign exerciseConcept

Crooze reservation management

The brief: help customers of Crooze, a fictional luxury car rental service, update their reservations after booking. I designed flows for adding and removing drivers, uploading driver's licenses, and changing delivery locations. The concept brings reservation details, map-based address selection, and clear update confirmations into a consistent visual language.

Edits open as in-line modals over the reservation. Adding a driver takes a license upload and a name, and removing one can be undone from the toast. The drop off map updates as the user types an address.

Reservation overview screen1 · ReservationAdd an additional driver modal with license front and back2 · Add driverMap modal for changing the drop off location3 · Change locationReservation with a confirmation toast after the location update4 · ConfirmationCrooze visual language: type, color, icons and imagery5 · Visual language
View the Figma exercise →
Oct 2025Kai CyberDesign exerciseConcept

Vulnerability findings, prioritized by AI

The brief: redesign a findings table so a security responder can quickly identify, prioritize and track vulnerabilities, and show how issues get resolved over time. With 436 findings and many false positives, responders don't know where to start. I put an AI priority score on every row, with a hover that explains why, and added a summary header with severity counts, MTTR and AI detection confidence. An open versus resolved trend line shows velocity and bottlenecks.

Redesigned findings table with KPI header, severity counts, resolution trend and AI priority tooltip1 · Findings tableDark version of the issue resolution trend chart2 · Resolution trend

Unfinished ideas

Concepts I have not resolved, kept visible because pretending a portfolio is only finished work is its own kind of dishonesty.

EV service experience

A self-initiated case study I have scoped but not built: a vehicle detects a fault, explains its severity and its confidence, and the owner is carried through triage, scheduling, and repair visibility without losing the thread.

Concept · self-initiated

Charging reliability

Availability is the number every app shows. Reliability is the number drivers actually need, and nobody surfaces it well. Early sketches only.

Sketch

Design systems for AI states

A component set for the states most systems leave undesigned: uncertain, degraded, refused, partially grounded. Pulled from Interpres, not yet generalized.

In progress

Cheap exploration first. Expensive rendering last. Never the other way around.

See the product work