Skip to content
Caldorin000/ Forging

Software studio · Web · Mobile · AI · Evaluation

Software, forged to ship.

6 engineers who design, build, and evaluate production software for teams that cannot afford a second attempt.

Start a project
See the work

01What we build

Four disciplines. One team that ships all of them.

Web, mobile, AI, and evaluation are not separate departments here. The same engineers who build the product build the harness that proves it works.

  1. 01Dashboards, marketplaces, SaaS.

    Web platforms

    Full-stack web products engineered for scale: typed APIs, real-time data, multi-tenant auth, and interfaces that stay fast under real traffic.

    • Product discovery & technical architecture
    • Design system and component library
    • Typed API layer and data modeling
    • CI/CD, observability, and handover docs
    • Next.js
    • React
    • TypeScript
    • Node
    • Postgres
    • Prisma
    • tRPC
  2. 02iOS and Android from one codebase.

    Mobile apps

    Cross-platform apps with native feel: offline-first sync, push, deep links, in-app purchases, and store submission handled end to end.

    • React Native / Expo or Flutter build
    • Offline data layer and background sync
    • App Store and Play Store release management
    • Crash analytics and release health monitoring
    • React Native
    • Expo
    • Flutter
    • Swift
    • Kotlin
    • Firebase
  3. 03LLM products that hold up in production.

    AI applications

    Retrieval pipelines, agents, and copilots with evaluation harnesses baked in, so quality is measured rather than assumed.

    • RAG and agentic workflows with guardrails
    • Prompt and model routing with cost controls
    • Offline eval suites and regression tracking
    • Fine-tuning and distillation when it pays off
    • Python
    • LangGraph
    • OpenAI
    • Anthropic
    • pgvector
    • Modal
    • vLLM
  4. 04Rubrics, red-teaming, and gold datasets.

    AI evaluation & data

    Structured human-plus-model evaluation for frontier and applied models: rubric design, adversarial testing, and high-precision annotation with QA.

    • Task and rubric design with calibration sets
    • Expert annotation across code, reasoning, and domains
    • Red-team and safety probes with reproducible reports
    • Inter-rater agreement and drift monitoring
    • Label Studio
    • Argilla
    • Python
    • DuckDB
    • Braintrust
    • Weights & Biases

02How we work

A process built to remove surprises.

01Week 0–1

Discover

We map the problem, the users, and the constraints, then write the architecture and estimate you will actually be able to hold us to.

  • Technical plan
  • Risk register
  • Fixed quote
02Week 1–2

Shape

Interfaces, data model, and integrations get designed together so nothing is discovered late. A working vertical slice proves the approach.

  • Design system
  • Data model
  • Vertical slice
03Sprints of 2 weeks

Build

Typed, tested, reviewed code in your repositories, with a demo every week and a written update you can forward to stakeholders.

  • Weekly demo
  • Written update
  • CI on every commit
04Continuous

Prove

Evaluation harnesses, load tests, and security checks run before release. For AI features, quality is a number on a chart, not an opinion.

  • Eval suite
  • Load & security report
  • Release gate
05Launch and beyond

Ship & run

Staged rollout, monitoring, and a handover that lets your team own the system. Stay on retainer or take it from here.

  • Runbooks
  • Dashboards
  • Handover session

03Selected work

Engagements, told as chapters.

Each project below is a representative engagement: anonymized, illustrative of the shape of work we deliver. Client-attributed case studies are shared on request.

04The team

6 engineers. No layers between you and them.

Senior generalists with deep lanes. Everyone on this page writes production code and talks to clients. Hover a shard to see who does what.

Six shards, one ring

05Ways to work together

Three shapes of engagement. Zero hidden line items.

01New products, MVPs, migrations

Scoped build

  • Fixed price and timeline
  • Architecture sprint first
  • Weekly demos
  • Full handover
02Ongoing product development

Embedded team

  • Two to six engineers
  • Your process and tools
  • Monthly retainer
  • Scale up or down monthly
03AI labs and applied ML teams

Evaluation program

  • Rubric design
  • Expert graders
  • Agreement metrics
  • Hourly or per-item

Not sure which fits?

Start with a two-week architecture sprint.

Book the sprint

06Questions

The things clients ask before they sign.

With a short discovery call, then a fixed-price architecture sprint of one to two weeks. You get a technical plan, estimate, and a working slice of the product before committing to a full build.

The six engineers on this page. We do not subcontract or resell. You will know who is on your project and talk to them directly.

Fixed-price for scoped builds, monthly retainers for ongoing product work, and hourly for evaluation and annotation programs. Every quote lists what is included and what is not.

We overlap at least four working hours with your team every day, run a shared channel, and send a written weekly update with demos. No surprises at the end of a sprint.

You do, from the first commit. Work happens in your repositories and cloud accounts wherever possible, under an agreement that assigns all IP to you.

Yes. We start with a two-week audit covering architecture, test coverage, security posture, and delivery risks, then propose a stabilization plan before adding features.

07Start a project

Tell us what needs to exist.

A few sentences is enough. You will hear back from an engineer, not a sales sequence, within one business day.

[email protected]

US / EU / APAC coverage

What are we building?

By sending you agree to our privacy notice.