The Parser Learns My Name Last

Parshv Patel

I am learning to write for machines

before I know how to speak for myself

formatting my ambitions into bullet points

spacing the margins to a quarter-inch

so the parser can find what I am

before any person does.

This is how you enter now

not through a door held open by a stranger

not through a referral phoned in over coffee

but through an upload box, a thank you screen

a silence that promises it has received you

and will be in touch.

I grew up watching my mother work hard

On the bus home, her shoulders held the shape of rooms

I thought I will not inherit this.

I will learn the language of the room at the top of the building.

But the room at the top of the building

now has a new gatekeeper

and it does not speak in the language of ambition.

It speaks in keywords, in parsed syntax,

in a scoring rubric no one publishes

and no one challenges

because challenging a number feels like arguing with a fact.

I have read the research.

The models trained on historical hires

reproduce the biases of everyone who hired before them.

They have learned what a good candidate looks like

from a century of records

that told a specific story

about who belongs in a corner office

and who belongs outside it, making the windows shine.

My high school was Title I.

Half the seniors I knew didn't apply to four-year colleges

because no one explained the application.

They were not unambitious.

They were unformatted.

They did not know to put Python before programming

data analysis before spreadsheets

the same skills in the wrong dialect

filtered to the bottom of the pile

before any human blinked.

I think about them constantly

especially now that I am learning to build the systems

that will do this faster

with higher confidence intervals

with less margin for the kind of error

that is actually a form of grace.

I believe in some of what this technology can do.

I have read about diagnostic tools

that caught cancer on a chest scan in a rural clinic

on the one day a week the radiologist appeared.

I have watched a translation model

give language to a grandmother speaking Gujarati

in a clinic waiting room in Fresno,

where before she had only her grandson's imperfect memory of the words

his small voice reaching for what she needed to say.

I have used tools

that compressed twelve hours of reading

into two hours of real understanding,

and I am not ashamed.

Access is not nothing.

For people who started with less

a tool that levels the library is a form of justice.

But justice requires intention

and intention requires a human in the room.

In warehouses right now

workers are scored every minute

pick rate, idle time, a bathroom break

registered as deviation from expected output.

A supervisor who used to use judgment

has been given a dashboard instead.

The dashboard cannot see

that her hands are shaking

because she drove two hours to make this shift

after the babysitter cancelled

after her second job ran over

after she sat in her car for twenty minutes in the dark

talking herself into going inside.

The dashboard sees, she is 4.3% below average.

It does not see the 4.3%.

It sees the deviation.

It recommends action.

Action is taken.

And somewhere in San Francisco,

the developer who built the dashboard

is proud of what he has optimized.

The word optimize is doing a lot of work these days.

So is efficiency

So is scale

These words mean different things

depending on whether you are the one being scaled.

I am studying to be the developer.

I am also the child of someone

who would have been on the dashboard.

I do not know how to hold both things at once

without one of them breaking.

So I carry them both

and I call it double consciousness

and I call it data science

and neither name fits exactly.

There is a company. I will not name it

that uses AI to schedule shifts for retail workers.

The software does not know

that a shift starting at 6 a.m.

and a shift ending at 11 p.m. two nights before

is called a clopening

is called exhaustion

is called the reason people quit

who then get flagged by the algorithm

for high turnover, for unreliability,

their instability is becoming a score

that follows them to the next application

like a letter of reference

they were never allowed to read.

The gig economy runs on this logic

you are independent until the app decides otherwise.

You can set your hours until the surge window closes.

You have the freedom of a function called at runtime

with no guarantee of the next call.

I graduated valedictorian of a school

and arrived at Berkeley on scholarships

that required essays about my hardship

which I wrote well enough to win

but which no one should ever have to write

just to access what others receive without application.

Now I am in AI4ALL program

studying the technologies

that will decide who gets to work at all.

I am the candidate the algorithm approved.

I am also watching it quietly deny

everyone who looks like where I came from.

I do not want to build systems

that replace human judgment.

I want to build systems that restore it.

I want to build the tool

that tells a hiring manager

here is someone who would have been filtered out

look at what the filter missed.

I want to build the audit, the appeal,

the architecture of reconsideration.

I want to build technology that treats a worker's schedule

the way a good manager would

with knowledge of the commute, the childcare window,

the second job, the preference for consecutive days.

Not because it is efficient to be humane.

Because efficiency built on someone else's exhaustion

is not efficiency.

It is transfer.

And we should call it what it is.

I am standing at the entrance to a future

that has already written most of the terms.

I am learning to negotiate.

I am learning which systems deserve to be improved

and which ones need to be rebuilt from the ground up

with different data, different goals,

different questions asked at the very start

not how do we screen faster

but how do we see better.

The future of work is not a dashboard.

The future of work is not a deviation score.

The future of work is not a parser

that learns my keywords

before it learns my name.

The future of work is a room

where someone holds the door without being asked

where the coffee is still warm

where a human being looks up and says

tell me something the resume couldn't.

And the answer takes longer than thirty seconds.

And someone stays to hear it through.

I am learning to build toward that room.

I am learning which machines to trust

and which to question loudly

in public, with data and poetry both.

I am carrying two inheritances

the language of algorithms

and the knowledge of what it means

to press your palms against the glass

and hold them there

until the window holds your shape.

Meet the Author

Parshv Patel is a sophomore at UC Berkeley studying Data Science with 4.00 GPA, driven by a belief that technology should create opportunity for those often overlooked. After moving from India during high school and overcoming financial, cultural, and academic barriers, he graduated valedictorian and now studies data science to build ethical, human-centered AI systems. An Amazon Future Engineer Scholar, AI4All Ignite Fellow, Greenhouse Scholar, and researcher, he explores machine learning, data storytelling, and algorithmic fairness. Outside academics, he mentors students, plays tennis, and enjoys turning curiosity into impact for communities facing barriers similar to those he experienced growing up himself firsthand.