Our work

What working with us looks like.

New studio, senior hands. Systems we've built handle 100k+ requests a day. Here's how we work — a live demo of one of the workflows we ship, and the four phases we run every engagement through.

100k+ requests / dayenterprise-grade systems shippedfixed-price scope · 48h proposal

Evidence strip

Selected work · anonymized

Customer support / legal ops

Seconds

from voice note to drafted reply

Whisper transcription · Claude drafting · RAG over approved replies

  • Python
  • FastAPI
  • Whisper
  • Claude
  • RAG

AI governance / compliance

ISO 42001

compliance-aligned controls

RBAC · multi-provider AI routing · evidence management

  • Next.js
  • Python
  • RBAC
  • Stripe

Legal (anonymized)

Instant

case files generated on demand

6 PDF + 3 Excel document types · cost calculations

  • Python
  • FastAPI
  • PDF
  • Excel

AI / SaaS

100k+

requests / day (production stack)

8 specialized analysis agents

  • Python
  • LangGraph
  • FastAPI
  • pgvector

Representative builds; some compose more than one project. Client details anonymized. Open-source work below is public and verifiable.

Open source

Public, verifiable work · github.com/eLyiN

codex-bridge

MCP server bridging AI agents to OpenAI's Codex CLI.

gemini-bridge

MCP server bridging AI agents to Google's Gemini CLI.

gemini-cli

Merged telemetry & activity-monitoring contributions to Google's official Gemini CLI.

Maintained MCP servers on PyPI, plus merged contributions to Google's Gemini CLI — all public and inspectable.

Live workflow demo · Inbound AI SDR

Claude-powered

Every inbound lead,
handled before you open Slack.

A LangGraph agent qualifies, enriches, and replies — before your SDR opens Slack. This is what one of our workflows looks like live.

  • Gmail

    source

  • Claude SDR

    agent

  • Salesforce

    output

  • Slack

    output

  • Gmail Reply

    output

Response
1.7s
Agents
4/4
Cost / run
$0.03

Illustrative values based on engagements we've shipped. Actual numbers depend on your traffic mix.

The four phases

Crystalline network visualization — discovery phase
01 / Discovery

Pair, scope, price

One or two weeks pairing with your team. We map integrations, surface hidden risks, and write a scope you can take anywhere — fixed price, no retainer required to start.

1–2 weeks

02 / Build

Ship fast, stay visible

Weekly demos. Every PR goes into your repo. You hold the keys from day one.

Kick-off
Sprint 1
Sprint 2
Demo
03 / Ship

Your cloud, your accounts

We deploy to infrastructure you own. No lock-in, no shared tenancy. Your keys, your logs, your SLAs.

  • Vercel
  • Docker
  • AWS
  • Azure
04 / Operate

Keep things running

Optional retainer to fix bugs, ship improvements and stay on-call when something breaks. Month-to-month. No multi-year contracts, no minimums.

Ask about retainers

Sound like what you need?

Tell us what you're trying to ship — we'll come back in 48h with a proposal.

Book a 30-min callSee what we ship