The Coolest Companies Using Haskell in Production

There is a persistent myth in software that Haskell is an academic language. A research toy. Something professors use to write papers and students use to pass exams. The myth is dead. Haskell runs in production at some of the most demanding, high-stakes, and genuinely interesting companies in the world. Here are the ones worth knowing about.

Meta: One Million Requests Per Second

Meta uses Haskell for Sigma, their anti-abuse platform. Every time someone posts, likes, comments, or messages on Facebook or Instagram, Sigma evaluates that action against abuse policies. The system processes over one million requests per second.

Meta previously used a custom language called FXL for this. They replaced it with Haskell and saw up to 3x performance improvements on certain requests. The migration produced Haxl, an open-source library for efficient concurrent data fetching that has become one of the most cited examples of Haskell solving real engineering problems at scale.

They also built Glean, a system for collecting and querying facts about source code. It powers developer tools across Meta: code search, IDE integration, code browsing. Meta is one of the largest sponsors of the Haskell Foundation. This is arguably the single largest-scale Haskell deployment in the world.

Tesla: Vehicle Firmware

This is the one that surprises people. Tesla uses Haskell for vehicle firmware development. Not for internal tooling or data pipelines. For the software that runs in the cars.

Tesla has been hiring Haskell engineers specifically for firmware roles. When the software in question controls a two-ton machine moving at highway speed, the choice of programming language is not a style preference. It is a safety decision. Tesla chose Haskell.

Mercury: 1.2 Million Lines of Haskell Banking

Mercury is a banking platform for startups. Their entire backend is Haskell. Not "some services are in Haskell." All of it. The monolith consists of approximately 10,000 modules containing around 1.2 million lines of Haskell code, making it one of the largest production Haskell codebases in existence.

Mercury has also given back to the ecosystem in serious ways. They built ghciwatch, a file-watching recompiler that loads 12x faster than HLS on large projects, and static-ls, a high-speed, low-memory language server designed for enterprise-sized Haskell codebases. When your codebase is 1.2 million lines, you either build better tooling or you suffer. Mercury chose to build.

Standard Chartered: $3 Billion in Haskell

Standard Chartered, one of the world's largest international banks, runs their entire Markets division on a Haskell dialect called Mu. The codebase is over 6.5 million lines. It handles deal valuation, risk analysis, and pricing across all asset classes. The Markets business line generated $3 billion in operating income in 2023.

What makes this remarkable is not just the scale. Over 100 people write functional code at Standard Chartered, and many of them are not developers. Financial modellers and traders write Mu to express business logic directly. The type system is not a barrier to adoption. It is the reason adoption works. When a trader writes a pricing function, the compiler catches the errors before the function touches real money.

Galois: When Correctness Is National Security

Galois was founded in 1999 specifically to apply functional programming to information assurance. They build formally verified systems for DARPA and the U.S. Department of Defense.

Their work includes SMACCMPilot, a hack-proof flight controller for unmanned aerial vehicles built under DARPA's HACMS program. They also won a $12.6 million DARPA contract for zero-knowledge proof research. Galois builds embedded domain-specific languages in Haskell that can mathematically prove systems behave correctly under all circumstances.

When the question is "can an adversary exploit this system?", the answer needs to be provably no. Not "we tested it and it seems fine." Provably no. That is what Haskell enables.

Anduril: Electronic Warfare

Anduril Industries is one of the fastest-growing defense technology companies in the world, valued at over $28 billion. They build autonomous systems, sensor networks, and electronic warfare platforms for the U.S. military and allied nations.

Anduril uses Haskell for developing correct, high-reliability software for controlling electronic warfare assets, integrated with their Lattice ecosystem. They hire embedded Haskell developers for this work. When you are building software that controls weapons systems, "it compiles and the tests pass" is not sufficient. You need stronger guarantees. Anduril chose the language that provides them.

Cardano / IOHK: Billions of Dollars on Haskell Types

The Cardano blockchain is written entirely in Haskell. Its smart contract language, Plinth (formerly Plutus Tx), lets developers write contracts in a subset of Haskell that compiles to Plutus Core. The blockchain has a market cap in the billions.

Input Output (IOHK), the company behind Cardano, chose Haskell explicitly because it provides mathematical certainty for cryptocurrency infrastructure handling billions of dollars. When a bug in your smart contract means someone loses their life savings, you want a type system that catches errors before deployment. IOHK is one of the largest Haskell employers in the world.

Juspay: 125 Million Transactions Per Day

Juspay processes payments across India. Their Haskell backend handles 125 million transactions per day with peak traffic of 5,000 transactions per second and 99.99% uptime. They specifically chose Haskell because its readability lets non-technical stakeholders verify business logic, and its strong typing helps achieve zero technical declines.

Juspay also performed one of the most ambitious language migrations in recent history: transpiling a large PureScript codebase to Haskell for performance benefits. The transpiled code is already handling production payments.

Wire: Encrypted Messaging

Wire provides end-to-end encrypted messaging for enterprises and governments. Their backend is developed primarily in Haskell with a microservices architecture. The server infrastructure has zero access to communication content by design.

When your product promise is that nobody, not even you, can read user messages, the implementation needs to be correct. Not "probably correct." Correct. Haskell's type system makes it possible to encode privacy guarantees at the type level, so violations become compile errors rather than security incidents.

Scarf: 2 Billion Downloads Per Day

Scarf tracks open-source package usage, processing over 2 billion downloads every day. The entire backend is Haskell. Founded by Avi Press, a board member of the Haskell Foundation, Scarf is a venture-funded startup that chose Haskell despite knowing the hiring challenges.

Avi wrote the famous "12 Reasons Why Haskell is a Terrible Choice for Startups (and why we picked it anyway)." The title is tongue-in-cheek. The argument is serious: when your system needs to be reliable at massive scale with a small team, Haskell's guarantees are not a luxury. They are a competitive advantage.

Chordify: Music Analysis in Haskell

Chordify is a music technology company that recognizes chords from audio. Their entire backend is Haskell. While music researchers develop chord recognition algorithms using deep neural networks in Python, Chordify runs all inference in production using Haskell.

This is not a typical use case. Running neural network inference in a functional language for a consumer music product is unexpected and genuinely cool. It works because Haskell is fast enough for real-time audio processing and reliable enough to serve millions of users.

GitHub: Code That Understands Code

GitHub built Semantic, an open-source tool for parsing, analyzing, and comparing source code across dozens of programming languages. They chose Haskell specifically because of its strength in working with abstract syntax trees: strong typing, lazy evaluation, purity, and rich possibilities for custom control flow.

The why-haskell.md document in the Semantic repository has become one of the most cited practical arguments for choosing Haskell. When the world's largest code hosting platform needs to build software that understands code, they reach for Haskell.

Supercede: Reinsurance Without Compromise

Supercede builds risk placement and analytics software for the reinsurance industry. About half the team are developers, all of whom spend most of their time writing Haskell, Elm, and Nix. Their CTO has stated he would "pick Haskell a thousand times over" if starting again.

Reinsurance is an industry where data quality is everything and mistakes are expensive. Supercede solves the unglamorous but high-value problem of cleaning and analyzing reinsurance data. Haskell's type system means data transformations are verified at compile time, not discovered in production when a policy is mispriced.

What These Companies Have in Common

Look at the list again. Anti-abuse detection at planetary scale. Vehicle firmware. Banking. National defense. Encrypted communications. Blockchain. Payment processing.

These are not companies that chose Haskell because it was trendy. They chose it because they operate in domains where bugs are not minor inconveniences. They are security breaches, financial losses, safety failures, or privacy violations. Every company on this list arrived at the same conclusion independently: when the cost of being wrong is high enough, the type system pays for itself.

The myth that Haskell is not a production language is not just wrong. It is the opposite of reality. The companies using Haskell are not doing it despite the difficulty. They are doing it because the difficulty is the point. The language filters for engineers who think carefully, and the type system catches the mistakes that slip through anyway.

If you want to work at companies like these, Typify is where you start.