Plaid Interview Questions — Page 2

Plaid·Software Engineer·Onsite - System Design / Architecture

May 2026

System design round at Plaid for a software engineer role. The prompt was basically 'build ChatGPT' which sounds exciting until you realize how many moving parts they actually want you to cover in 45 minutes.

Design a conversational AI assistant similar to ChatGPT. Your design should support multi-turn conversations with history, token-by-token streaming responses, high availability under peak load, safety controls like toxicity filtering and PII handling, and optional tool use such as calling a search API or calculator. Walk through the high-level architecture, key components, storage choices, scaling strategy, and how you'd evaluate quality and safety in production.

“I started with the happy path, which in hindsight was the wrong move.”

View Post

Plaid·Frontend Engineer·Onsite - System Design / Architecture

Jun 2026

System design round at Plaid for a frontend engineer role, focused entirely on designing something like Plaid Link itself. Pretty meta. The question was dense and covered a lot of ground, from iframe embedding all the way to token storage and PII isolation.

Design a third-party embeddable widget (similar to Plaid Link) that lets users connect their bank accounts to a partner application. Cover the iframe embed model, the API surface, OAuth and credential flows, the data model, security considerations, and how the partner app fetches account data server-to-server.
How would you handle step-up authentication and MFA within the bank credential flow, including how redirects are managed between the bank, your backend, and the embedded widget?
What security measures would you put in place around the iframe postMessage communication and how would you prevent a malicious host page from intercepting or spoofing messages?
How would you design the data model for users, linked accounts (items), institutions, and access tokens, and how do you isolate PII across partner tenants?

“This is basically six questions in a trenchcoat.”

View Post

Plaid·Software Engineer·Technical Phone Screen

Jun 2026

Plaid SWE interview with a coupon-discount coding problem that escalated pretty quickly from a single coupon to multi-coupon stacking with a bunch of edge cases to handle. Clean code, tests, and complexity discussion all expected.

You have a shopping cart where each item has a category and a price, plus a list of coupons. First, implement applying a single coupon (fixed discount or percentage off a category) and compute the final total. Then extend it to handle multiple coupons with stacking rules you define. Make sure to handle: discounts exceeding the item price (floor at zero), the same category targeted by multiple coupons, and percentage discounts over 100%. Write tests and discuss time complexity.

“The single-coupon phase felt manageable but I underestimated how much design thinking the multi-coupon part required.”

View Post

Plaid·Machine Learning Engineer·Onsite - System Design / Architecture

Jun 2026

System design round at Plaid for an MLE role, focused entirely on building a fraud detection system for a fintech data-connectivity platform. The question was massive and kept branching into follow-ups I wasn't fully ready for. Left feeling like I covered maybe 70% of what they wanted.

Design an end-to-end ML system for detecting fraud across the full user lifecycle on a fintech data-connectivity platform, covering bank-account linking, identity verification, and suspicious transaction behavior. The system needs near-real-time scoring and graded actions (allow, step-up, review, block) with a low false-positive rate.
A fraud ring is sharing a single device across 50 newly created accounts, each linking a different bank. How would your graph features and velocity counters detect this, and how fast?
Your offline PR-AUC improved after a model update but online fraud loss didn't drop. What are the likely causes and how would you diagnose it?
How do you evaluate and set thresholds for a model when its own blocks are suppressing the labels you'd use to measure it? What strategy keeps your measurement unbiased?
The model is flagging a legitimate immigrant population at a higher rate due to IP and geolocation features. How do you detect this disparate impact and fix it without gutting recall?

“This is the main question and it's enormous.”

View Post

Plaid·Software Engineer·Technical Phone Screen

Jun 2026

Plaid SWE interview with a meaty data engineering problem involving routing number deduplication and conflict resolution across multiple JSON sources. The problem had a lot of moving parts and felt more like a mini system design than a pure coding question.

You're given bank routing data from multiple JSON sources. Each source has records mapping routing numbers to bank names, but the same bank can appear under different aliases, the same routing number can appear multiple times within a source, and different sources can disagree on which bank a routing number belongs to. Write a function that normalizes bank names using an alias map, produces a final routing-to-canonical-bank mapping, resolves conflicts using majority vote across sources with a tiebreaker based on source priority, and outputs a conflict report listing which routing numbers had disagreements and how the vote went.

“This one took me a few minutes to even parse what was being asked.”

View Post

Plaid·Software Engineer·Onsite - System Design / Architecture

Jun 2026

Plaid software engineer interview that was essentially a full system design presentation. You pick a project you owned and walk through the whole thing on slides, from problem framing to postmortem. More prep-intensive than I expected.

Walk through a significant project you owned end-to-end using a slide deck, covering the problem, goals, stakeholders, constraints, architecture, data model, APIs, design decisions, trade-offs, performance and scalability, testing and rollout, metrics, incidents, and lessons learned.

“This is basically a 30-45 minute presentation and they will poke at every layer.”

View Post

Plaid·Product Manager·Onsite - Product Sense / Strategy

Jun 2026

Plaid PM interview with a product strategy question about helping Netflix roll out ACH payments in the US. Pretty focused case, one question but it had a lot of layers once you started pulling on it.

You're helping Netflix introduce ACH as a payment option in the US. How would you design a test to decide whether it should be rolled out broadly?

“My first instinct was to jump straight into experiment design but I realized pretty fast I hadn't defined what success even looked like for Netflix here.”

View Post

Plaid·Software Engineer·Technical Phone Screen

Jun 2026

Plaid data engineering interview that threw a deliberately messy, under-specified data problem at me. The core task was building a fuzzy bank-name mapping pipeline, and the ambiguity was very much the point. Walked away thinking I should have asked more clarifying questions upfront instead of diving straight into code.

Write a function that normalizes a raw bank name string into a canonical form, handling varying capitalization, punctuation, abbreviations, and corporate suffixes like 'N.A.', 'Inc.', and 'Corp.'.
Given cleaned bank name data with a non-1:1 mapping between internal and external bank IDs, write a fuzzy mapping function that returns all plausible external bank IDs for each internal bank ID rather than a single match.

“This part felt manageable.”

View Post

Plaid·Software Engineer·Onsite - System Design / Architecture

Jul 2026

Plaid data engineering round built around a take-home assignment, where you sit down in a locally configured environment and write SQL while talking through your design decisions out loud. The interviewer then picks apart every choice you made until something breaks.

Walk through how you designed the pipeline from raw source data through to the final analytical layer. Why did you structure it the way you did?
Star schema or snowflake for this model, and what drove that choice?
How would you handle incremental loads for this pipeline? What's your strategy and what are the failure modes?
What happens to your model if events arrive out of order or timestamps have timezone inconsistencies?
Your pipeline has duplicate rows coming in from the source. How does your model handle that, and where does deduplication live?
The source schema changes and a new column gets added, or an existing column gets renamed. How does your pipeline not break?
How do you handle deletes in the source system? Does your model support hard deletes, and what are the implications?
There are NULLs in your join keys. What does your model do with those rows?

“This was basically the whole interview in one question.”

View Post

Plaid·Software Engineer·Onsite - Behavioral / Leadership

Jul 2026

Plaid software engineer interview with a presentation-heavy round that basically turns into a 15-minute interrogation of a project you pick. More prep-intensive than I expected, especially the follow-up questions which went pretty deep.

Walk us through a recent project you led or contributed to significantly. Cover the problem, constraints, timeline, how you estimated work, and how you managed risk.
How did you delegate work and coordinate with your teammates during this project?
How did your collaboration with your manager differ from how you worked with a peer IC on this project?
What were the key technical or product trade-offs you made, and how did you decide between them?
How did you handle scope creep during the project?
Describe a time you had to re-prioritize under a tight deadline. What got cut and how did you make that call?
How did you communicate and align with other teams during the project?
What outcomes and success metrics did you use to evaluate the project, and did you hit them?
What would you do differently if you ran this project again?
Walk us through a conflict that came up during the project and how you resolved it.

“This is the core of the whole round.”

View Post