Architecture Review.

A written assessment, typically 15–25 pages, covering:

• System map. The actual state of your data, retrieval, classification, and agent infrastructure, drawn as a diagram you can hand to your team.
• Findings. What's working, what's brittle, and what's quietly limiting your AI roadmap. Prioritized by leverage, not by alarm level.
• Recommendations. Concrete architectural moves. Not “consider building a semantic layer.” More like: “your supplier model should canonicalize on tax ID, not name; here's the migration path and the three risks.”
• A 90-day plan. What to do first, whether with me or without.

• ·Canonical data models, taxonomies, and entity resolution
• ·Retrieval architecture: vector, graph, hybrid, internal search
• ·Classification systems and the workflows built on top of them
• ·Agentic workflows, tool use, and approval patterns
• ·Evaluation, observability, and the loop between them
• ·Where the system will break when scope or scale changes

1. 01

   Kickoff call (60 min)

   Context, goals, who to talk to, what to prepare.

2. 02

   Working sessions

   Three to five over one week, with the people who own the systems. Engineers, data leads, AI/ML practitioners. Not just leadership.

3. 03

   Write week

   I produce the draft. You review. I revise once.

4. 04

   Readout (60 min)

   Walk through findings with your team and leadership.

Two weeks, end to end. Remote.

Fixed-fee, scoped per engagement. 50% on signing, 50% on delivery. If the scope is wrong before we start, I'll say so and refund the deposit.

You're a good fit if you're:

• ·Operating in classification-heavy, data-fragmented, or regulated environments where generic AI patterns are starting to show their limits.
• ·Past the pilot phase, but unsure why production reliability is harder than the demos suggested.
• ·Considering a larger AI investment and want a senior, independent read before committing.

You're not a good fit if you're earlier than that (at “should we try AI?” or “should we build an internal chatbot?”). I'll tell you that on the first call and point you somewhere useful.

Common follow-on engagements include data foundation work (canonical modeling, entity resolution, data preparation for AI) and platform-specific build-outs, most often on Databricks.

Email fbaig@htsmcp.com with a few sentences about what you're building. I respond within two business days.

Have a system you're stuck on? Talk it through with me in 30 minutes →