AI expertise.Human judgment.

aikeep humans in the loop.
youwhere humans stay in charge.
scroll
// what we do

Verticals

[01]
Knowledge Infrastructure

Engram

Contextual memory as a service.

4.2M
facts indexed / wk
92%
recall @ k=5
<40ms
p95 query latency

We stopped repeating decisions in Slack. The memory travels.

Platform eng lead, early access
  • Managed ingest & retrieval API
  • 12 integrations (Slack, GitHub, Linear, …)
  • Team-aware knowledge graphs
  • SOC 2 compliant, single-tenant option
Read the Engram case →
[02]
Hospitality

Sobremesa

The CRM that remembers your guests.

+27%
repeat visit rate
3.1×
booking conversion
9.4
operator NPS

We finally know our regulars before they walk in. The kitchen knows too.

Owner-operator, 60-seat restaurant
  • Guest profiles & preference tracking
  • Reservation management & waitlist
  • Automated follow-ups & re-engagement
  • Kitchen & front-of-house sync
Read the Sobremesa case →
// open source & research

Labs

featured project

engram engine

An open-source memory engine for AI systems. Engram catches what gets said, links it to what was already known, and brings it back when it's useful.

example.ts
import { Engram } from "@humanlayerdo/engram"

const engine = new Engram({
  memory: "persistent",
  context: "structured",
})

await engine.ingest(conversation)
const recall = await engine.query("last decision on auth")
View on GitHub →
research interests

Contextual memory

Knowledge that survives the next context switch.

Human-in-the-loop patterns

AI proposes. The human decides. The design problem is where the seam goes.

Agent orchestration

How agents hand off work without losing the thread.

// capabilities

Team

Engineering

Full-stack systems we run in production. Cloud, APIs, observability, on-call. The unglamorous parts that keep things up.

AI / ML

Applied AI: integrations, fine-tuning, retrieval, agents. Evals over vibes.

Design

Product and interface design that takes complexity seriously. AI systems should be easier to read than the prompt that built them.

Strategy

Technical advisory for teams adopting AI. Where to start, what to skip, what to ship first.

How we engage

Sprint
2–4 weeks

Discover and prototype. A working artifact by week three.

Build
2–6 months

An embedded squad ships production software next to yours.

Advisory
ongoing

Fractional technical leadership for teams mid-rollout.

Have a project in mind?

> Tell us about it.