We treat automation as engineering — typed inputs, deterministic boundaries, retries, observability, and evaluations. Not just an LLM behind a button.

PIPELINE / invoice-agent.v3

LIVE

TRIGGER

Event ingest

CLASSIFY

LLM router

AGENT

Tool-using loop

VALIDATE

Schema + policy

DATABASE

Postgres · pgvector

NOTIFY

Slack · Email · Webhook

p5082ms

p95142ms

success99.4%

events/day24,891

SELECTED WORK

Projects that
shipped, scaled
and stayed in production.

View all case studies

PROJECT / 01

IN PRODUCTION

HREZ·SaaS · HR Tech

A modern HRIS built for growing teams — from records to payroll in one system.

Next.jsTypeScriptPostgreSQLPrismaTailwindAWS

Read case study

PROBLEM

Fast-growing SMBs were running HR out of spreadsheets, shared drives and legacy payroll tools. Employee records, leave, attendance and payroll lived in four different places, with no single source of truth for people data.

SOLUTION

HREZ unifies employee records, org structure, attendance, leave, performance and payroll under a single multi-tenant SaaS platform. Every entity is typed and event-sourced, so audits, exports and integrations reuse the same contracts.

manual HR hours

-72%

tools replaced

uptime

99.98%

PROJECT / 02

IN PRODUCTION

RAG Agent·AI Infrastructure

A production-grade retrieval agent teams drop into their existing chatbot.

TypeScriptNode.jsPostgreSQLpgvectorAnthropicOpenAI

Read case study

PROBLEM

Teams shipping chatbots kept rebuilding the same retrieval stack — chunking, embeddings, re-ranking, grounding, evals — and kept getting the same production issues: hallucinated answers, stale sources, and no observability into why an answer went wrong.

SOLUTION

A hosted RAG agent with a typed API and a drop-in JavaScript client. Ingest documents from URLs, files or connectors; ask questions from any chatbot with a single call; get grounded answers with cited sources, streamed to the UI.

p95 answer latency

< 900ms

grounded answer rate

94%

time to integrate

1 day

PROJECT / 03

SHIPPING

Arkeen·Consumer · Mobile

A calm, private baby life journal — for the moments parents actually want to keep.

React NativeTypeScriptExpoSupabasePostgresS3

Read case study

PROBLEM

Most baby tracking apps are cluttered, ad-supported and treat every tap like a data point. Parents wanted something quiet — a private place to log the day and keep the memories that will matter years from now.

SOLUTION

Arkeen is a mobile-first journal that combines fast logging (feeds, sleep, diapers) with a soft memory layer (photos, first words, milestones), packaged in a calm, editorial interface.

week-4 retention (beta)

94%

cold-start to first log

< 2s

third-party trackers

PROCESS

A delivery model
built like software,
not like consulting.

Six stages, one team, one shared repository. Each stage produces a deliverable that survives the engagement — not a slide deck.

01
Discover
Understand the operating model and the constraints.
02
Architect
Design the system, data model and integrations.
03
Design
Build the product experience end-to-end.
04
Build
Implement the platform in typed, tested, deployable slices.
05
Launch
Deploy, instrument and observe in production.
06
Optimize
Measure outcomes and iterate against real usage.

INTEGRATIONS

An integration
for every seam.

We connect to the tools your team already uses — no rip-and-replace. Data flows through typed contracts with retries and observability.

AWS

Vercel

Cloudflare

Supabase

PostgreSQL

Redis

Temporal

OpenAI

Anthropic

Pinecone

Weaviate

Stripe

Slack

HubSpot

Salesforce

Google

Microsoft

Notion

Linear

SIGNAL

Numbers, not narrative.

PRODUCTS SHIPPED

INDUSTRIES SERVED

0.00%

UPTIME TARGET

0.0M+

PROCESSES AUTOMATED

WHAT TEAMS SAY

A few notes from
the people who ran it.

“Roei rebuilt the operational core of our platform in one quarter. We are still running on their architecture two years later.

Marta Bell

Head of Ops · Northstar Capital

“They approached it like engineers who happen to be good at product — not the reverse.

Daniel Ng

CTO · Halo Studio

“The clearest thinking about AI in production that we've worked with. Nothing shipped that we couldn't debug.

Priya Rao

VP Engineering · Atlas Logistics

START A CONVERSATION

Ready to build
something real?

Tell us what you're trying to build. We'll respond with a short technical read, not a sales deck.

Start a conversation Explore our work

RESPONSE

< 24h

DISCOVERY

1 week

FIRST DEPLOY

3–5 weeks

We architectand build thesystems behind modern business.

Everythingnecessary toship software.

Product Strategy

UX / UI Design

Web Applications

SaaS Platforms

AI Systems

API & Integrations

Software shouldremove complexity,not add to it.

Architecture first

Product thinking

Automation

Measurable

Software mappedto how the workactually happens.

HR Tech

AI Infrastructure

Consumer Mobile

Financial Services

Professional Services

Logistics

Every workflowis a small,verifiable system.

Projects thatshipped, scaledand stayed in production.

A modern HRIS built for growing teams — from records to payroll in one system.

A production-grade retrieval agent teams drop into their existing chatbot.

A calm, private baby life journal — for the moments parents actually want to keep.

A delivery modelbuilt like software,not like consulting.

Discover

Architect

Design

Build

Launch

Optimize

An integrationfor every seam.

Numbers, not narrative.

A few notes fromthe people who ran it.

Ready to buildsomething real?

Everything
necessary to
ship software.

Software should
remove complexity,
not add to it.

Software mapped
to how the work
actually happens.

Every workflow
is a small,
verifiable system.

Projects that
shipped, scaled
and stayed in production.

A delivery model
built like software,
not like consulting.

An integration
for every seam.

A few notes from
the people who ran it.

Ready to build
something real?