Full-stack · AI systems · Product engineering
I’m Jared, a full-stack software engineer. I build products end to end—from model and tool integrations through APIs, data, interfaces, infrastructure, and deployment.
Selected work · 2023—2026
01 Agent orchestration
Verified software delivery with coding agents
A distributed delivery system that coordinates bounded agent tasks, isolated runner work, verification evidence, and a human review boundary before publication.
02 Developer tools
A local-first coding-agent platform
A provider-neutral agent runtime shared across CLI, terminal, editor, and desktop clients, with workspace context and visible controls over tool access.
03 AI research systems
Evidence-first research and writing
A research workspace that turns a question into objectives, MCP-assisted source discovery, reviewable evidence, and papers whose claims stay linked to their sources.
04 Production SaaS
Truth-preserving resume tailoring
A resume workflow that uses AI to tailor a candidate's own experience to a role, while keeping edits reviewable and unsupported claims out of the result.
05 Production SaaS
B2B lead discovery and outreach workflows
A lead-generation SaaS that combines public business discovery with persistent jobs, lead management, filtering, exports, and outreach integrations.
How I approach the work
Build around the decisions and handoffs a person has to make, not a demo that stops after the first prompt.
Separate UI, APIs, persistence, model providers, tools, and permissions so each part can be reasoned about.
Keep outputs reviewable: citations, diffs, verification results, and clear failure states belong in the product.
Engineering notes
Multiple persistence layers can work well in one SaaS product when each piece of data has an authoritative owner, a recovery story, and a deliberate lifecycle.
A usage ledger, idempotent payment events, and explicit charging rules keep paid AI workflows consistent when requests fail or webhooks retry.
Connect each requirement to a versioned check and its evidence so a green test run tells a reviewer what the agent actually proved.
