life2 is a boutique AI engineering firm. Agents that do real work, customer-experience AI, Meta platform integrations, and the AWS infrastructure underneath.
every run traced, scored against evals, and observable in your own AWS account
Four practices. One team.
AI, LLMs, machine learning, agentic systems. We take on the whole build: the intelligence, the integration, and the infrastructure it runs on.
cap/01
Agentic AI systems
Agents that plan, call tools, verify their own results, and hand off cleanly when they should. We design the loop, the guardrails, and the evaluation harness that proves it behaves before it meets a customer.
tool use orchestration evals scored on real traces guardrails typed, tested
cap/02
Customer-experience AI
Assistants your customers actually reach: support that resolves cases, routing that lands with the right person, follow-ups that happen. Grounded in your data and measured on resolution.
support resolution flows voice + chat surfaces grounding your sources
cap/03
Meta platform integrations
WhatsApp, Messenger, and Instagram as product surfaces. Business messaging, commerce flows, and AI agents that answer where your customers already spend their day.
WhatsApp Business Platform Messenger + IG messaging agents on every thread
cap/04
AWS infrastructure
The platform under the intelligence: Bedrock, serverless compute, data pipelines, identity, observability. Infrastructure as code from the first commit, designed to be operated by your team.
Bedrock model platform serverless + containers IaC reviewed, versioned
Software like infrastructure: built once, run for years.
Architecture first. Decisions documented before code, tradeoffs visible.
Evals before launch. An AI system ships with the harness that measures it.
Embedded with your team. We work alongside engineering and product, not over them.
Yours to operate. Runs in your cloud account, documented so your team can keep it running.
The drawing comes first. The system is built to match it.
We build our own products the same way.
The practices above are not theory. They run our own product line every day.
Product
Paredra
An AI that listens beside you in meetings, builds a private knowledge graph of your work and life, and answers from your own history. Built end to end on the stack we sell: agents, Bedrock, serverless AWS.
More products are in design on the same architecture-first line. We announce them when they are real, not before.
The mark, proposed.
Mark B carried the last verdict and stays the base. Three alternatives evolve it toward the corrected identity: an engineering firm that builds in layers, modules, and multiples. Each is shown flat, filled, and in horizontal lockup. The decision remains a founder verdict.
B. The two courses
The base. Built work laid in rows: the firm above, the product line beneath.
B1. The stack
Three layers of a running system: infrastructure, model, agent. The working layer is green.
B2. The module
The two courses joined by a trace. Systems built as parts, shipped connected.
B3. The exponent
Life, squared. The course of work, raised by the small green block: software that multiplies.
All four hold the brand laws: flat, filled, horizontal lockup, no serif. One green element per mark. Companion Paredra marks are staged on the paredra.ai preview.