Case studies / LLM products
LLM products
Prompting, assistants and copilots: shipping a product built on an LLM.
Approach
13 write-ups
Tag:
-
DropboxNewLLM products
How AI-assisted development helped us build the new Dropbox experience
Problem Engineers needed to add complex new features to web and desktop quickly, but development was slow and fragmented.
Solution They used AI to generate code, implement features, and run tests, letting engineers review and approve the work, enabling fast cross‑platform delivery.
Read on dropbox.techDetails -
VercelLLM products
How Interfaze serves 50 million requests a month on Vercel
Problem Customers needed reliable, deterministic AI results for tasks like document OCR and web scraping, but single model providers were flaky and non‑deterministic.
Solution They built an AI gateway that routes requests across multiple model providers, runs model‑generated code in Vercel’s sandbox, and falls back automatically when a provider fails.
Read on vercel.comDetails -
ShopifyLLM products
Cross-store product clustering across millions of stores
Problem Different merchants list the same product in many ways, making it hard to show a unified view.
Solution They created tiered matching models, graph partitioning, and a learned merge step to cluster billions of listings across stores into a single identifier.
Read on shopify.engineeringDetails -
GoogleLLM products
Unlocking Earth AI’s planetary geospatial foundation models for global public health
Problem Public health data is slow, sparse, and split by borders, slowing response to outbreaks.
Solution Created a planetary geospatial foundation model that turns satellite, mobility, and search signals into monthly location embeddings, which can be dropped into existing health models without extra training.
Embeddings matched or improved conventional inputs across many disease tasks
Research behind this, on AI Papers DailyBasin2Vec: Learning Basin-Centric Embeddings for Scalable Hydrological ModelingRead on research.googleDetails -
NVIDIALLM products
Build Local AI Apps with C++ and NVIDIA TensorRT RTX Samples
Problem Developers needed a way to run AI models locally on Windows, Linux, and ARM without custom runtimes.
Solution Created open‑source C++ samples that export models to ONNX, then use ONNX Runtime with TensorRT for fast GPU‑accelerated inference across platforms.
Read on developer.nvidia.comDetails -
YelpLLM products
The making of the Yelp Assistant UI
Problem Mobile app updates are slow, so UI changes take days to reach users.
Solution Created a server‑driven UI where the backend sends layout and content, letting Yelp Assistant’s chat UI be updated instantly without new app releases.
Read on engineeringblog.yelp.comDetails -
ShopifyLLM products
Helix: The internal tool powering our Shopify app's native migration
Problem Migrating a large React Native app to native Swift/Kotlin was slow and error‑prone.
Solution Created Helix, a tool that splits migration into tiny checkpoints with strict review gates, letting LLMs generate code step‑by‑step while engineers guide it.
App rebuilt and published in 12 weeks
Read on shopify.engineeringDetails -
Hugging FaceLLM products
Granite 4.2 LLMs: How They're Built
Problem We needed a language model that could reason step‑by‑step and use external tools when needed.
Solution We built Granite 4.2 by pre‑training a dense decoder, fine‑tuning on chain‑of‑thought data, then applying multi‑stage reinforcement learning that teaches tool use and reasoning, releasing 3 B, 8 B, and 30 B versions.
Read on huggingface.coDetails -
ExpediaLLM products
Focus on the Feature, Not the Fixture: GenAI powered GraphQL mocks
Problem Developers spent hours creating and updating hand‑written GraphQL mock data that quickly became outdated.
Solution They created mockql‑rs, a CLI that reads a GraphQL query with @mock tags, checks the schema, and asks an LLM to fill in only the annotated fields, producing schema‑valid mock responses on the fly.
Read on medium.comDetails -
TripadvisorLLM products
Taming a Million Tickets: How a Small Team Used AI to Modernize an Overgrown Jira Instance at Scale
Problem Over a million Jira tickets were scattered across 245 projects with inconsistent fields and workflows.
Solution They built an LLM‑powered system that read ticket data, identified patterns, and automatically migrated tickets to a unified configuration.
Read on medium.comDetails -
ReplitLLM products
Unlocking a New Way to Build Enterprise Data Apps with Microsoft Fabric
Problem Teams spent weeks setting up authentication, APIs, and governance to turn data into production apps.
Solution They created Rayfin, an open‑source SDK that turns a simple description into backend code, adds security and compliance, and deploys the app directly to Microsoft Fabric.
Read on replit.comDetails -
JioHotstarLLM products
Beyond Imitation: Teaching AI to Speak Like a Character in Hindi
Problem Hindi speech synthesis lacked emotion and consistency at production scale.
Solution Created a modular LLM‑based TTS system with Hindi data curation, language‑aware preprocessing, and production‑focused evaluation to generate expressive, stable speech.
Read on blog.hotstar.comDetails -
The New York TimesLLM products
How The New York Times is scaling Unit Test Coverage using AI Tools
Problem Test coverage lagged behind rapid feature development, causing extra manual testing.
Solution Engineers used generative AI agents to automatically write unit tests, then humans reviewed and refined them, quickly raising coverage on critical code.
Read on open.nytimes.comDetails
No write-ups match these filters.