Commercial context: Opportunity referred by Prokopto; account and deal are Resbyte (Brainforge ↔ Resbyte), not a Prokopto resale.
Summary (from Granola notes export)
Resbyte — product overview
- Resbyte builds a data-driven ad personalization platform for restaurant marketing.
- Three layers of personalization:
- Content (e.g. vegetarian vs. chicken burger).
- Real-time behavior (e.g. Friday dinner orders trigger Friday dinner ads).
- Customer value (e.g. high-LTV customers get aggressive couponing).
- Workflow: marketer builds a simple 4–7 step logic for one profile → AI applies it to millions of profiles.
- All ad assets generated upfront (overnight/weekends) → library of up to 1M+ ads ready before campaigns run; system picks the right asset at the right time; no on-the-fly generation.
- Resbyte owns the targeting logic entirely — Meta/Google used only as last-mile delivery; cohort-level instructions (e.g. “these IDs see this ad at 9:15”) rather than letting Meta optimize spend; belief that Meta/Google algorithms optimize for themselves, not the client.
Current tech stack and challenges
- Hosted on AWS; logic and data processing in Snowflake; nothing substantial built in AWS beyond hosting.
- Two immediate problems:
- Gen AI for ad unit assembly — model to stitch brand images into fixed templates at spec across format variations; no suitable model yet (not generating brand imagery from scratch — library + template assembly).
- Intelligent agent layer — personalization logic is basic; needs to get smarter.
- Scale: thousands of ads today → path to tens of thousands+ for larger brands.
- Long-term vision: autonomous system reading real-time consumer behavior, optimizing 24/7 with minimal manual input (audience ID, zip saturation, product prefs, competitors → gen AI execution).
Next steps
- Uttam (Brainforge): High-level architecture for AWS-based image generation pipeline (open-source vs licensed e.g. Flux on Bedrock), options, core steps, Snowflake integration touchpoints, sample output types — share with Pt for review.
- Pt (Resbyte): Example ad templates + sample campaign output / customer journey to scope architecture.
- Pt / Naren: Share long-term architecture documents for a follow-up conversation.
- Timeline: Active campaign through May — partial improvement in May valuable; open to 3–5 month roadmap post-funding.
Full verbatim transcript
Open the meeting Brainforge & Resbyte (2026-05-01) in Granola using ID 37de78dc-dfa0-47ad-be8c-13b178a9908d, or re-export via Granola MCP (get_meeting_transcript). This vault file intentionally keeps the summary as the durable index; raw ASR is long and best viewed in Granola.