--- schema: 1 type: video title: 'AI tools for Forward Deployed Engineering — Vasuman Moza, Varick Agents' description: >- A customer spent five million dollars and five years migrating to SAP, and has zero appetite to rip anything out again. That constraint is the whole design at Varick Agents: instead of asking an enterprise to migrate, you drop forward deployed agents on top of the systems they already run. Vasuman Moza's version of the role maps how a department actually works today, the reconciliations between a purchase order and an invoice, the handoffs nobody documented, and then automates those processes end to end. It answers the stat everyone cites, that most AI projects never reach production, by starting from the customer's real workflow instead of a generic tool. Making one forward deployed engineer that productive takes tooling of its own. Varick builds a spec from the raw material of an engagement, the granola notes and Slack threads, then lets the engineer shape the workflow with Claude or Codex against a single source of truth that can just live in Postgres. The harder problem is context: frontier models are surprisingly bad at traversing a messy enterprise and knowing that person A and person B are the same entity, so Varick post trains its own models to extract the right context and strip the redundancy, and only then does the agent run autonomously. Every engagement starts by finding the bottleneck and grows out from there. Speaker info: - https://x.com/vasuman - https://www.linkedin.com/in/vasumanmoza/ Timestamps: 0:00 - Introduction: forward deployed agents 1:55 - The question: what can an agent actually do? 3:27 - What a forward deployed agent is 4:44 - Automating a department end to end 6:14 - Why enterprises need deployed engineers, no migrations 8:30 - Hiring the top 1 percent 10:51 - The platform and department wide ROI 11:53 - Demo: tools for forward deployed engineers 13:07 - Turning notes into a workflow spec 14:22 - Engineering workflows with a source of truth 16:53 - Post training models to extract context 18:37 - Where this leaves us --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/l0FLhNqBOic-ai-tools-for-forward-deployed-engineering Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- canonical: 'https://liquidgold.cc/watch/l0FLhNqBOic' markdown: 'https://liquidgold.cc/watch/l0FLhNqBOic.md' llms: 'https://liquidgold.cc/watch/l0FLhNqBOic/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T06:29:55.365Z' language: null children: [] videoId: l0FLhNqBOic durationSeconds: 1223 publishedAt: '2026-07-28T21:00:06.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 23 endSeconds: 120 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=23' heldReason: rights - ordinal: 1 text: null startSeconds: 130 endSeconds: 210 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=130' heldReason: rights - ordinal: 2 text: null startSeconds: 220 endSeconds: 287 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=220' heldReason: rights - ordinal: 3 text: null startSeconds: 297 endSeconds: 381 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=297' heldReason: rights - ordinal: 4 text: null startSeconds: 392 endSeconds: 515 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=392' heldReason: rights - ordinal: 5 text: null startSeconds: 392 endSeconds: 515 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=392' heldReason: rights - ordinal: 6 text: null startSeconds: 526 endSeconds: 655 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=526' heldReason: rights - ordinal: 7 text: null startSeconds: 655 endSeconds: 667 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=655' heldReason: rights - ordinal: 8 text: null startSeconds: 800 endSeconds: 869 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=800' heldReason: rights - ordinal: 9 text: null startSeconds: 800 endSeconds: 869 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=800' heldReason: rights - ordinal: 10 text: null startSeconds: 881 endSeconds: 1015 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=881' heldReason: rights - ordinal: 11 text: null startSeconds: 1015 endSeconds: 1027 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=1015' heldReason: rights - ordinal: 12 text: null startSeconds: 1027 endSeconds: 1117 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=1027' heldReason: rights - ordinal: 13 text: null startSeconds: 881 endSeconds: 1015 watch: 'https://liquidgold.cc/watch/l0FLhNqBOic?t=881' heldReason: rights --- # AI tools for Forward Deployed Engineering — Vasuman Moza, Varick Agents A customer spent five million dollars and five years migrating to SAP, and has zero appetite to rip anything out again. That constraint is the whole design at Varick Agents: instead of asking an enterprise to migrate, you drop forward deployed agents on top of the systems they already run. Vasuman Moza's version of the role maps how a department actually works today, the reconciliations between a purchase order and an invoice, the handoffs nobody documented, and then automates those processes end to end. It answers the stat everyone cites, that most AI projects never reach production, by starting from the customer's real workflow instead of a generic tool. Making one forward deployed engineer that productive takes tooling of its own. Varick builds a spec from the raw material of an engagement, the granola notes and Slack threads, then lets the engineer shape the workflow with Claude or Codex against a single source of truth that can just live in Postgres. The harder problem is context: frontier models are surprisingly bad at traversing a messy enterprise and knowing that person A and person B are the same entity, so Varick post trains its own models to extract the right context and strip the redundancy, and only then does the agent run autonomously. Every engagement starts by finding the bottleneck and grows out from there. Speaker info: - https://x.com/vasuman - https://www.linkedin.com/in/vasumanmoza/ Timestamps: 0:00 - Introduction: forward deployed agents 1:55 - The question: what can an agent actually do? 3:27 - What a forward deployed agent is 4:44 - Automating a department end to end 6:14 - Why enterprises need deployed engineers, no migrations 8:30 - Hiring the top 1 percent 10:51 - The platform and department wide ROI 11:53 - Demo: tools for forward deployed engineers 13:07 - Turning notes into a workflow spec 14:22 - Engineering workflows with a source of truth 16:53 - Post training models to extract context 18:37 - Where this leaves us --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/l0FLhNqBOic-ai-tools-for-forward-deployed-engineering Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- Canonical: https://liquidgold.cc/watch/l0FLhNqBOic · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.