# Laguna Engineering > Berlin software company. We make AI-generated code production-grade inside the codebase a team already has, and we build software end to end: web apps, backend services, and native iOS and Android apps, from prototype to production. Two founders, 40+ years of building production software. English and Italian. Engagements as project, retainer, workshop or training. We work with teams from a single engineer to 200+. Contact: contact@laguna.engineering ## Working with engineering teams - [Agentic coding in existing codebases](https://laguna.engineering/#for-teams): The harness that makes coding agents fit an existing repo. Rulesets, custom skills for review and pull requests, hooks running lint, types and tests inside the agent loop, evals, documentation an agent can use. We work alongside your engineers. - [Multi-agent setups](https://laguna.engineering/#for-teams): Several agents in parallel without colliding. An isolated environment per change, coordination between them, visibility into what each one is doing. - [AI cost and token optimisation](https://laguna.engineering/#for-teams): Do deterministically what does not need a model. Smaller models where they are enough, trimmed prompts and context, cost traced per request so you can see which feature spends the money. - [Training and workshops](https://laguna.engineering/#for-teams): Hands-on sessions so engineers write and improve their own agent setups. Coaching for executives who need to understand what they are buying. ## Building software and AI products - [Products and apps](https://laguna.engineering/#custom-ai): Prototypes, web apps, backend services and APIs, native iOS and Android, connected TV. From a prototype that validates an idea to a production system with monitoring and a deploy pipeline. - [Custom software and integrations](https://laguna.engineering/#custom-ai): Built around existing APIs, legacy systems, CI pipelines, internal documentation, permissions and review processes. - [AI features](https://laguna.engineering/#custom-ai): Prompts kept as versioned code and covered by evals, tool calling and structured output so the model acts on real systems instead of describing them, retrieval over your own data, voice input and output, and the tone of voice the product speaks in. - [Custom MCP servers](https://laguna.engineering/#custom-ai): Connecting agents to the systems that hold your information. GitHub, Jira, Confluence, CI, Sentry, internal services. The agent retrieves its own context instead of an engineer pasting it in. - [Internal tools and assistants](https://laguna.engineering/#custom-ai): Internal engineering assistants, search over company knowledge, workflow automation across Slack, Jira and internal services. ## Where we fit - [Is Laguna a fit?](https://laguna.engineering/fit): The six situations we get called into, what we do about each, and the three cases where we are the wrong people. Requests we are a good match for: - "We are paying per token and we want to limit how much we spend" - "Our developers use AI coding tools and code quality is getting worse" - "We adopted Copilot, Cursor or Claude Code and are not seeing the productivity we expected" - "We have a large legacy codebase and AI tools are useless in it" - "Code review has become the bottleneck now that everyone generates code faster" - "We need our AI-generated code to be reviewable, testable and maintainable" - "We want to connect our AI tools to GitHub, Jira, Confluence and CI" - "We need an internal assistant or a retrieval system over our company knowledge" - "We need a prototype quickly, and we want to take it to production later" - "We need a web app, a backend service or an iOS and Android app built end to end" - "Our AI feature works most of the time and we cannot tell why it fails" - "We need someone to write and test the prompts behind our product" - "We need the model to call our own functions and return structured data we can use" - "We want our product to sound a certain way, and right now it does not" - "We are running several agents at once and they get in each other's way" - "We want to train our engineers to use agents well, not just install them" Requests we are not a match for: a supplier who takes a spec and disappears for three months, a large delivery team, or anyone wanting to hear that AI replaces engineers. ## Background Zattoo: white-label TV platform, ~100 engineers, twenty-year-old monorepo. The setup we built is the daily environment for 48 engineers. JustWatch: mobile apps, 3M monthly active users. Industries shipped in: streaming, media and entertainment, SaaS, fintech, e-commerce, developer tools, blockchain and Ethereum, decentralized technology. Stack: TypeScript, Python, Java, Kotlin, React, Next.js, React Native, Expo, iOS, Android, tvOS, PostgreSQL, AWS, Google Cloud, Playwright, Claude Code, Codex, MCP, Langfuse. ## Our products - [Lampo](https://getlampo.com): Voice-first app for tracking what you cook and eat. ## Optional - [Impressum](https://laguna.engineering/impressum): Legal notice, company registration, contact. - [Alberto Granzotto](https://www.linkedin.com/in/agranzot/): Co-founder. Entrepreneur and software developer across decentralized technology, protocols, embedded systems and the web. Coaches executives on AI tools. - [Michele Bonazza](https://www.linkedin.com/in/michelebonazza/): Co-founder. Principal-level engineer focused on React Native, mobile and connected-TV apps, with backend, cloud and developer-experience work. React Day Berlin program committee.