Engineering service
AI platforms & agent orchestration
Multi-agent workflows with planning, memory, typed tool calls, and human checkpoints — from creative-strategy platforms to assistant co-pilots.
- Typical slice
- 4–12 week milestones
- Engagement
- Hands-on build + architecture
- Core stack
- Mastra · Temporal · Langfuse
In depth
I design and ship agent systems where planning, tool use, and memory are first-class — not a single prompt wrapped around an API. That means clear boundaries for what an agent can do, durable workflows for long-running work, and surfaces (Slack, dashboards, extensions) where humans stay in the loop when it matters.
Production examples include Parker AI’s Mastra-based ideation and reporting stack with Temporal behind ingestion and heavy jobs, and Get Magic’s assistant agents with MCP-style tool gateways so execution stays auditable and scoped.
What you get
Agent architecture: plans, memory scopes, tool graphs, and human checkpoints
Durable workflows for ingestion, fan-out, and long-running jobs
Typed MCP-style tools and adapters to your APIs, ads platforms, and data stores
Product surfaces aligned to how your team works (Slack, Next.js, extensions)
Tracing and eval hooks so every release is measurable
How we work
Map the workflow
Identify decision points, risky actions, and where humans must approve.
Design the system
Tool contracts, memory model, and orchestration (sync vs durable).
Build in vertical slices
Each slice is shippable, traced, and demoable to stakeholders.
Harden for production
Budgets, rate limits, retries, and least-privilege tool scopes.
Handoff
Runbooks and patterns your engineers can extend without guesswork.
Examples & past work
Outcomes you can expect
- Ship agent runtimes where tool calls are typed, traced, and bounded by production rails.
- Orchestrate long-running work on durable workflow engines (e.g. Temporal) alongside real-time product surfaces.
- Deliver Slack-, dashboard-, and extension-first UX so outputs land where teams already work.
Questions, answered
Do you only work with Mastra and Temporal?
No — those are representative stacks from shipped work. I match your runtime (or help you choose) based on durability, team familiarity, and operational constraints.
How do you decide what should be an agent vs a traditional service?
Agents earn their place when the task needs planning, tools, or memory across steps. If it is deterministic and cheap to hard-code, we keep it boring on purpose.
What does a first milestone look like?
Usually one end-to-end workflow — traced, gated, and demoable — plus a clear diagram of how tools and memory evolve in the next milestones.
Book a free 30-minute discovery call to investigate your work and needs
I will map constraints, risks, and a practical first milestone — whether that is agents, retrieval, ingestion, extensions, or full-stack SaaS delivery.
Other services
Vector + relational retrieval, semantic chunking, and re-ranking so LLM outputs stay grounded when catalogues and documents get large.
Schedulers, retries, and normalised pipelines from ads APIs, social platforms, and internal services — built for scale and observability.
Chrome extensions and sidepanel experiences with in-browser RAG and real actions via tool gateways — not another detached chat tab.