Skip to content
View TAM-DS's full-sized avatar

Block or report TAM-DS

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. 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
TAM-DS/README.md

Tracy Anne Griffin Manning. AI Solutions Engineering & Technical Sales | Enterprise Discovery to Deployment | Advisory & Governance | Business Value First. Understand the workflow. Then build the AI.

AI Solutions Engineering & Technical Sales | Enterprise Discovery to Deployment | Advisory & Governance | Business Value First
Based in Austin; open to San Antonio, Dallas–Fort Worth, and Houston; in-office or hybrid; open to meaningful business travel.
Website · Email · LinkedIn · X

I build enterprise AI for decisions where money, risk, and accountability matter—connecting financial-services experience with AI architecture, infrastructure economics, and controlled execution.

Current: Founder, Apex AI|ML
Prior: Minotaur Consulting · Office of the CIO · Wall Street


Featured — FirstSync: governed customer onboarding, deployed

FirstSync · live demo — no login or API key

An AI agent that onboards a new customer's Salesforce data into a B2B product. Claude inspects the org and proposes every field mapping; a deterministic dry-run shows exactly what would load before anything is written; a person approves — and approval binds to a hash of the proposal, so a later change invalidates it.

The clearest expression of the operating principle this profile is built on, applied to a problem I first met in 2014 leading a third-party risk onboarding delivery — where the technology was never the hard part; coordination was. On a held-out org the prompt was never tuned against, the mapping scored 29/29 with perfectly calibrated questions.

Reference implementation with no customers; the demo runs on precomputed model proposals, so it needs no API key.


What I bring — three areas of impact

Business value first. Inspectable engineering behind every claim.

01 / AI Strategy & Infrastructure Economics

Make the economics inspectable. Translate AI architecture into cost, performance, and risk trade-offs that executives can evaluate before committing capital.

Evidence: Review the financial assumptions.
Published FinOps figures are synthetic and modeled; career outcomes are separate.

02 / AI Governance, Cyber Risk & Audit Evidence

Make accountability visible. Build explicit authorization boundaries, accountable ownership, and traceable evidence for human review.

  • VANGUARD-Guard — AI inventory, risk assessment, and ownership workflows.
  • AEGIS-NHI — Connected investigations and modeled financial exposure.
  • AEGIS Evidence — Replayable workpapers and SHA-256-verifiable archives.
  • GuardRails-MCP — Prompt-regression checks and human release-review evidence.

Evidence: Inspect an evidence-linked decision brief.
Reference workflows and synthetic cases; framework mappings are not certification.

03 / Capital Markets, Energy & Reliable Execution

Keep research and execution accountable. Connect financial and operational context with research tools, supply-chain decision support, and controlled execution.

Evidence: Inspect worker-recovery and retry tests.
Historical research and simulated execution; no live trading or investment-return claims.

Enterprise integration in practice

SignalDesk connects evidence-linked sales research with bounded, read-only Salesforce access, opt-out checks, and human-authorized outreach decisions. Its Salesforce GET-only implementation was selected by the XERJ team for its public code-reference corpus. Developer-org and simulated-HTTP evidence; no automated CRM writes or outreach.

Prior career outcomes: $1.2M+ annual cloud savings · 30% less manual reporting · platforms supporting 10,000+ daily users. These are professional-history outcomes, separate from portfolio simulations.

Operating principle: The model can propose. The system determines what is authorized.


Data & platform foundations

Multi-cloud (AWS, GCP) and Databricks lakehouse engineering — governed data pipelines, Unity Catalog, and infrastructure economics an executive can inspect.

Repository Focus
Databricks AI-Ready Lakehouse Governed Spark / PySpark pipelines, Unity Catalog, Delta Lake, read-only MCP boundaries
Idempotent Data Engineering Foundation One-command path from raw CSV to market intelligence with true idempotency
FinOps for M&A Infrastructure DD Modeled multi-cloud infrastructure risk scenarios (illustrative savings, not a client result)
AI Infrastructure Intelligence Modeled multi-cloud GPU cost scenarios

Multi-model engineering & runtime architecture

The model is a component. The architecture is the product.

I assess OpenAI GPT/Codex, Anthropic Claude, Google Gemini, and xAI Grok according to workload fit: reasoning and structured output, code and tool-use workflows, context requirements, retrieval, latency, cost, and failure behavior. Model evaluation is not the same as a deployed four-provider orchestration layer.

The governing design principle remains independent of provider: explicit tool grants, deterministic policy checks, source-linked evidence, bounded execution, and human accountability. See SignalDesk, AEGIS-NHI, and Agent Foundry for inspectable examples of these boundaries.


Open-source reference engineering

XERJ Corpus Hub selection (2026) — SignalDesk Salesforce GET-only integration

The XERJ team selected a bounded Salesforce GET-only preview implementation from SignalDesk for inclusion in its public code-reference corpus. The selected work focuses on validation, fail-closed behavior, opt-out checks, and deterministic restrictions around enterprise API access.

Engineering lesson: The model may propose; the system determines what is authorized.

Selection for indexing is external recognition of an implementation—not certification, formal peer review, a universal reference standard, proof of agent adoption, or measured token savings. A direct corpus citation can be added when publicly verified.


Supporting engineering depth

Agent Foundry separates capability, identity, and runtime grants, with a documented AWS DEV proof. Supporting boundary examples include Monster Desk, Monster Light, and AEGIS Analyst.


Try a demo

Open the interactive BALLAST walkthrough — no login, installation, or API key. Compare synthetic energy and semiconductor scenarios and replay denied versus approved mitigation outcomes.

This hosted browser walkthrough uses precomputed results from the Python engine. It is not a hosted Python backend, authenticated approval service, or live trading system. Inspect the exporter and CI results.

Run the full applications locally

Each repository documents its dependencies and entry points.

Gradio

  • BALLAST — supply-chain control tower (energy routes, semiconductor concentration, approval-gated mitigation)
  • AEGIS Analyst — SOC analyst console with policy checks and sandboxed actions

Streamlit trading workbenches (historical daily data + Python metrics + bounded CrewAI; agents have no order tools)

# typical Streamlit entry
python -m streamlit run app.py

What I bring

AI Solutions Engineering & Technical Sales | Enterprise Discovery to Deployment | Advisory & Governance | Business Value First

Verifiable shipped code. Python, cloud architecture, and MCP integration. Bounded tool authority and model evaluation. Founder-level ownership. Translation from an executive business problem to a reviewed technical decision.

These are reference implementations. They demonstrate architecture and controls a reviewer can inspect—not claims of paying customers or enterprise production deployment.

Pinned Loading

  1. firstsync firstsync Public

    When a B2B company signs an enterprise customer, go-live stalls for weeks on integration work. FirstSync gives the work to Claude.

    Python

  2. signaldesk-acquisition-intelligence signaldesk-acquisition-intelligence Public

    B2B sales-development research and review application. OpenAI powers optional live company research and evidence-assisted drafting. The application enforces human review.

    Python 1

  3. agent-foundry agent-foundry Public

    Most agent frameworks hand the model standing keys and hope. Agent Foundry separates capability, identity, and runtime authority — and proves the boundary on AWS with scoped, short-lived grants.

    Python 1

  4. databricks-ai-ready-lakehouse databricks-ai-ready-lakehouse Public

    Data pipelines aren't AI-ready until they are governed: this Databricks framework powered by Apache Spark and PySpark uses Unity Catalog and Delta Lake to serve verified, audit-ready data products …

    Python

  5. ai-infra-command-tower-finops ai-infra-command-tower-finops Public

    Repository models that decision loop across AWS and Google Cloud. Includes AI/ML usage telemetry so infrastructure economics can be examined from both a finance and platform-engineering perspective.

    Python 1

  6. agentic-rag-voice-intelligence-platform agentic-rag-voice-intelligence-platform Public

    What does agentic AI actually do for revenue? For a real client: +32% traffic, +18% sales YoY. An agentic RAG knowledge & voice platform on n8n + OpenAI + ElevenLabs + Supabase. Names changed.