Caltech Essays 2026-27, With Examples: The Complete Guide

What Caltech Is Looking For

Caltech evaluates applicants on three things: intellectual curiosity, scholarly character, and drive. Curiosity is whether you chase ideas past the point they're required of you. Scholarly character is how you learn, especially when there isn't an easy answer. Drive is whether you've actually done something with a STEM interest, not just held one.

Area of Interest

Why did you choose your proposed area of interest? (150-200 words)

What to do

• Trace your interest back to a specific moment, question, or piece of work that started it, not a general love of science.

• Get technical. Caltech faculty read these, so naming the actual mechanism, method, or concept you're interested in reads as sincere in a way vague enthusiasm doesn't.

• Show where the interest is headed, a question you still want answered, not just where it started.

What to avoid

• Don't write an origin story with no technical content in it. "I've always loved biology" doesn't tell a faculty reader anything about how you think.

Example: My interest in plant genetics started with a mistake. I was helping my uncle graft a Valencia orange branch onto a sour orange rootstock in his backyard, and I grafted the scion upside down. It survived anyway, badly, with growth twisted toward the ground before it corrected itself over two seasons. That accident made me curious about what the graft union actually is: not just glue between two plants, but a live boundary where vascular tissue from two genotypes has to reconnect and start communicating hormonally. I started reading about how rootstock choice changes scion gene expression, specifically how rootstocks can alter fruit sugar content and disease resistance in the top-grafted variety without changing its DNA. That's what pulled me toward genetics specifically, not just agriculture: the idea that a plant's environment, including the roots it's literally been given, can shift which genes get expressed without touching the genome itself. I want to study how epigenetic signaling crosses a graft union, and whether that mechanism could be deliberately used to make drought-resistant rootstock lines carry that resilience into unrelated scion varieties.

Scholarly Character

Caltech asks you to choose one of two options for this essay. Both are getting at the same thing, how you learn when the answer isn't easy, but they come at it from different angles.

Option A: Collaboration

Tell us about a time your learning in STEM was shaped by another person or group. (150-200 words)

What to do

• Be specific about what changed in your thinking, not just that someone helped you.

• Show the collaboration going both directions if it did. Caltech is asking how you learn with others, not just how others taught you.

What to avoid

• Don't make this about a group project logistics story. It needs to be about a specific shift in understanding.

Example: I spent most of junior year convinced that elevators in tall buildings used a single motor per car, scaled up from the elevator in my apartment building. A classmate in my physics research group, Wren, corrected me mid-argument by pointing out that a cable-driven system for a fifty-story building would need a counterweight close to the car's own mass just to keep motor torque manageable, and that traction, not brute motor power, is what actually limits how fast you can move a car safely. I pushed back, because it felt like a technicality, until she walked me through the math on why counterweight ratio changes braking distance in an emergency stop. That correction changed how I approach problems generally: I had been treating "how does X move" as a question about power source, when for anything large enough, the real question is almost always about force distribution and safety margins instead. We ended up building a small scale model together to test how counterweight ratio affected stopping distance under sudden load, and I learned that the fastest way to learn something in STEM is to argue with someone who already understands the part you're wrong about.

Option B: Process

Tell us about a time your approach to a STEM problem mattered as much as the outcome. (150-200 words)

What to do

• Show a specific decision point in your process, not just the fact that you persisted.

• Answer the part of the prompt that asks what you'd still defend about your process regardless of outcome. That's the part that separates this from a generic perseverance essay.

What to avoid

• Don't resolve the essay with a clean success. Caltech is asking about process specifically because it wants to know how you think when the outcome is still uncertain.

Example: I spent six weeks trying to determine tooth wear patterns in a set of donated deer molars for an independent research project, using a method I designed myself instead of the standard mesowear scoring most of the published papers used. Mesowear scores cusp shape by eye, which felt too subjective for what I wanted to measure, so I built a simple profilometry setup instead, using a dial indicator to trace cusp height loss across the tooth surface in quarter-millimeter increments. It was slower, and my first data set was noisy enough that I couldn't tell if the pattern I was seeing was real or just measurement error from an unsteady hand. I still don't have publishable results. But I would defend the decision to build my own measurement method rather than eyeballing a mesowear score, because the process taught me exactly where the measurement error was coming from, my grip pressure varying trial to trial, which is something a subjective visual score would never have let me diagnose. A wrong number I understand is more useful to me than a plausible number I don't.

Scientific Drive

Caltech asks you to respond to two of three prompts here. All three examples below are fully drafted so you can see how each angle plays out.

Option A: Learning

Take this opportunity to nerd out about a STEM rabbit hole. (150-200 words)

Example: I fell into a rabbit hole about elevator dispatch algorithms after getting stuck for four minutes in a crowded lobby watching two elevators arrive at the same floor within seconds of each other while people upstairs were still waiting. That seemed like an obvious inefficiency, so I looked into how dispatch actually works, and learned that older systems use simple first-come, first-served logic, while modern destination dispatch systems have riders enter their floor before boarding so the system can group people heading to nearby floors into the same car before anyone steps in. The part that surprised me was learning that destination dispatch isn't purely about speed. It's solving a version of the bin-packing problem in real time, where the "bins" are elevator cars with limited capacity and the "items" are passengers with destination constraints and patience thresholds that decay the longer they wait. I started building a basic simulation in Python that models wait time under both dispatch strategies with randomized passenger arrival, and I'm still tuning the passenger patience function, since real complaints spike well before real wait times do.

Option B: Pursuing

Tell us about a STEM question that has held your attention over time. (150-200 words)

Example: I've spent almost two years trying to figure out why a set of X-rayed teeth from a local dental office's discarded imaging archive showed inconsistent enamel density readings across supposedly identical exposure settings. It started as a question about equipment calibration, but the pattern only showed up in certain patients, not others, which ruled out a simple machine error. I read into how enamel density on radiographs is affected by mineralization gradients that form during childhood, meaning two visually identical teeth can radiograph differently based on developmental history the imaging can't otherwise show. I've been using ImageJ to quantify grayscale density across a de-identified sample of the archived films, looking for a correlation between density variance and reported patient age at the time of the scan. What I still don't understand is whether the variance I'm measuring reflects genuine mineralization differences or an artifact of exposure angle that I haven't fully controlled for, and that's the question I want to keep working on: not what the pattern is, but whether I'm actually measuring the thing I think I'm measuring.

Option C: Making

Tell us about something you created, tested, or redesigned. (150-200 words)

Example: After my county's small farm insurance co-op flagged a spike in claims tied to late blight, I built a basic computer vision model to flag early blight symptoms on potato leaves from phone photos, since most of the affected farmers were reporting the disease only after it had already spread across a field. I trained the model on a public leaf disease image set using edge detection to isolate lesion boundaries, then compared lesion shape irregularity against known early-stage blight patterns, since early lesions tend to have more diffuse, water-soaked edges than the mature, well-defined lesions most training sets are built around. My first version had a false negative rate around forty percent on early-stage images, because the model had mostly learned to recognize mature lesions. I retrained it on a smaller set of early-stage-only images I annotated myself, which cut the false negative rate roughly in half. It's still not reliable enough for a farmer to trust over their own judgment, and that gap, between a model that performs well on a test set and one a working farmer would actually use, is the part of this project I think about most now.

Just for Fun

What is something you would be excited to do, share, teach, make, start, or contribute as part of the Caltech community? (100-150 words)

What to do

• Let this one actually be fun. It doesn't need to connect to your academic interest, and it usually reads better when it doesn't.

• Be concrete about what you'd actually do, not just a personality trait you'd bring.

Example: I want to start a refrigerator roulette night in whatever dorm I end up in, where everyone contributes one random ingredient from their own fridge or the dining hall and someone has to cook a full dish from whatever the pile turns out to be. I've been doing a version of this with my own family since I was thirteen, mostly because my dad kept buying ingredients for recipes he never finished, and it became a running joke to force him to use the half a jar of capers or lone stalk of lemongrass sitting in the door. It's turned me into a genuinely fearless cook, since you can't be picky about ingredients when the whole point is working with whatever's there. I'd bring that same refusal to waste a weird ingredient to Caltech, and probably talk someone into eating a caper and lemongrass stir fry along the way.

Optional: Academic Context

This section is optional and exists to give context, not to make a case for yourself. If your school doesn't offer AP or IB courses, if you didn't have access to a particular lab or research opportunity, or if something in your transcript needs a brief explanation, this is the place for it. Keep it factual and short. It isn't another essay.

Working With Momentum

Caltech's essays get read by faculty, and it shows in how specific and technical a strong answer needs to be. Our team of advisors, all recent graduates of schools like Harvard, Brown, Columbia, Cornell, Dartmouth, Duke, Georgetown, and UChicago, can help you find the right level of detail for each prompt. Reach out to get started.