Bernstein
Open-source multi-agent orchestrator. 254 stars, ~33K PyPI downloads/month, 37 CLI agent adapters, Apache 2.0.
Consulting · 2026
Multi-agent orchestration, deployed reliably.
I help engineering teams run AI-agent systems before they blow up in production.
Forward-deployed engineering, on a swarm. Drop Bernstein into a client repo and you get a multi-agent crew with file-based state, per-agent credential scoping, and an HMAC-signed audit trail. Runs on whichever CLI agents the client already trusts.
Best fit: engineering-first orgs, 5–50 engineers using AI agents daily, Series A–C.
Credentials
Built, not blogged.
Aporia · TeamInternet
Sole engineer behind the marketing-ops AI platform (2026–present). Real production traffic, agentic operations, three-tier tool calling.
Prior depth
Storage and infrastructure engineering, R&D at VTB, a decade of systems work before AI was the interesting part.
The problem
The agent zoo, once it stops being cute.
A 2026 engineering team running Claude Code, Cursor, Codex CLI, or several of these at once tends to hit the same failure shape within a few months.
Agent drift
Agents drop tool calls after 30 turns. Different CLIs use different prompt formats, context windows, tool-call quirks. Nothing composes.
No audit trail
An agent commits code on Sunday night. On Monday no one can say why it made that decision. Replay is impossible. Postmortems are guesswork.
No quality gates
Agents merge code without running tests. Lint, types, contract checks all bypassable. Failures land in main and someone notices on Tuesday.
How I help
Three engagement shapes.
Pricing depends on scope, team size, and how deep the integration goes. Send a short email and I'll reply with rates and a proposed shape.
Level 1 · 1 week
Orchestration Assessment
Diagnose the current agent stack. Document failure modes from the past three months. Recommend an orchestration architecture (Bernstein, LangGraph, or homegrown). The diagnostic is honest about what does not need replacing. Deliverable: 15–25 page assessment and an executive readout.
Pricing in the email
Level 2 · 2–4 weeks
Forward-deployed engineering setup
Parachute onto your repo and stand up an AI engineering crew on Bernstein. Wrap 3–5 of your existing agent tools as adapters. Configure quality gates, HMAC-chained audit trail, janitor verification. Train 1–3 engineers on day-to-day operation.
Pricing in the email
Level 3 · 2–3 months
Agentic Platform Architecture
Custom agentic platform for a specific workflow: coding org, internal vertical, or operator surface. Includes custom adapters for your internal tools and 30 days of post-launch support.
Contact me for scope and pricing
Fit
What this is and isn't.
Narrow blade. Works well in the right hand, not at all in the wrong one.
Good fit
- Engineering-first orgs where the CTO or VP Eng owns the decision.
- 5–50 engineers using AI coding agents daily; reliability is now an operational issue.
- Series A–C startups with $2–50M ARR. Infrastructure, DevTools, B2B SaaS, developer platforms.
- Teams that have read the writing and reach out themselves.
Not a fit
- "Just want to cut engineering costs with AI": wrong problem, wrong person.
- 90-day enterprise procurement and an NDA maze for a small engagement.
- Consumer-facing AI app development; full-stack builds with AI features bolted on.
- Prompt-engineering workshops, model fine-tuning, RAG-as-a-service.
Open inquiry
If your team is running AI agents in production and the wheels are wobbling, email me. A short paragraph on what you're running and where it hurts is enough to start.
I read every email and usually reply within one business day.