Skip to content
View hishamalward's full-sized avatar

Block or report hishamalward

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
hishamalward/README.md

Hisham Al-Ward

AI platform and automation engineer in Toronto. I build agentic AI and automation systems for enterprises, and I measure whether they were worth building.

Three things I care about

  • Agentic governance: AI agents writing production code behind spec gates, blast-radius guardrails, and quality gates that make violations mechanical rather than discouraged.
  • Automation with receipts: value-stream baselines, honest ROI models, and measured-vs-projected kept strictly separate.
  • LLM engineering: per-task model routing, prompt evals calibrated against human labels, and cost engineering. Listenality's enrichment pipeline runs at ~$0.36 per 1,000 tracks because the evals said it could.

Public work (MIT, each with a docs/learning/how-it-works tour for whoever owns it next)

Repo What it does
evalmine Scores a model change on your own tasks: pairwise LLM judge with position swap, Cohen's kappa against your labels, schema and execution checks, latency, cost from a pinned price table. Refuses to headline a win-rate its judge cannot defend. Three-tool MCP server so an agent can run the evals mid-task.
mcpclerk A governance proxy for MCP servers: per-tool allowlist (deny by default), human approval for write-class tools, quotas, secret redaction before logging, and a hash-chained audit log that verifies.
agentkeel A framework for shipping production code with AI coding agents: work priced by size, spec-first, four gates with named owners, blast radius bounded by Claude Code hooks.
agent-slots Per-agent isolation for parallel coding agents: one integer derives a worktree, a database, two ports, and a job-queue schema.
toilscan Claude Code plugin that scans Git history for recurring developer toil and recommends the smallest automation that removes it. Every finding cites its commits.

Background: A decade+ across networking, cloud consulting, and security automation. Azure Solutions Architect Expert, AWS Solutions Architect, Terraform, CCNP.

Toronto · LinkedIn

Pinned Loading

  1. toilscan toilscan Public

    Analyze Git repositories for recurring developer toil and recommend the smallest useful automation.

    Python

  2. agent-slots agent-slots Public

    Per-agent isolation for parallel AI coding agents: one integer derives a worktree, a database, ports, and a job-queue schema

    Shell

  3. agentkeel agentkeel Public

    A framework for shipping production code with AI coding agents: work priced by size, spec-first, four gates with named owners, blast radius bounded by hooks

    Python

  4. mcpclerk mcpclerk Public

    A governance proxy for MCP servers: per-tool allowlist, human approval for write-class tools, quotas, secret redaction, and a hash-chained audit log

    Python

  5. evalmine evalmine Public

    Scores a model change against your own tasks: pairwise LLM-judge win-rates calibrated to your labels, schema-pass, execution checks, latency, cost. Refuses to headline a number its judge cannot def…

    Python