I help B2B SaaS companies right-size their data infrastructure and ship AI workflows they can actually trust in production.
Less Stack. More Signal.
Most companies hit a wall where their data stack — the Snowflake contracts, the Fivetran subscriptions, the Hightouch syncs — stops making sense for the workload they actually have. And most teams building AI-powered features get stuck in the "last mile": the prototype works, but nobody can answer the risk questions stakeholders need answered before they'll approve a production launch.
Flynn Data Services works on both problems, fractionally.
Track 1 — Data Infrastructure
Right-sizing data stacks, reducing tooling costs, and building pipelines that work. Recent results at Cybrary:
- 92% reduction in search infrastructure costs
- 66% reduction in CRM licensing costs
- 96% improvement in CRM data latency
Track 2 — AI Workflow Architecture
Structured risk assessment, human checkpoint design, evaluation harnesses, and cost modeling for teams building AI-powered products. The goal: shifting the stakeholder conversation from "is this safe?" to "here's exactly what could go wrong and what we're doing about it."
| Repo | What it does |
|---|---|
| fds-snowflake-governance-blueprint | Config-driven Snowflake RBAC governance framework |
| fds-snowflake-cost-audit | 10-query workload audit with Evidence BI report output |
| fds-ai-workflow-audit | AI workflow risk assessment across 5 dimensions + cost calculator |
| bi-in-a-box | DuckDB + dlt + Evidence — enterprise analytics at startup cost |
Languages: Python (expert), SQL (expert), R (intermediate)
Data: Snowflake, Airflow, dbt, DuckDB, dlt, Evidence BI
Cloud: GCP (Cloud Run, Pub/Sub, Cloud Build), AWS (Lambda, S3, Step Functions)
MarTech/RevOps: HubSpot, Salesforce, Hightouch, Google Ads, GTM
AI/Agents: Claude API, MCP, Supabase + pgvector, agentic workflow design
Currently taking select engagements. Fixed-fee projects, monthly retainers, and hour blocks available.