Inspiration
Every small business owner we talked to described the same exhausting loop: write the caption, resize the image for five different platforms, remember to post at the right time, check if the ad spend is working, and somehow also find time to run the actual business. Marketing agencies charge thousands of dollars a month to do this, and most solo founders and small teams simply go without — posting inconsistently, guessing at what content works, and leaving money on the table with campaigns nobody has time to optimize. We wanted to build the tool we wished existed: something that could think like a marketing team, not just a scheduling calendar.
What it does
Digilize is an AI-powered digital marketing platform that acts as an on-demand marketing department. Tell it your brand, upload a few documents about your business, and it:
- Writes and designs content — generates on-brand captions, images, and short-form video for every major platform
- Publishes on autopilot — schedules and posts directly to Facebook, Instagram, LinkedIn, Twitter/X, and TikTok
- Grounds itself in your brand — a RAG pipeline lets it pull real facts from your uploaded docs instead of hallucinating product details
- Runs the ad side too — helps plan and manage campaigns across Google Ads and Meta Ads, backed by real SEO keyword research
- Reports back — a single analytics dashboard shows what's actually working across every channel
- Talks to you like a strategist — a built-in marketing chat assistant answers questions and takes action, instead of just chatting
How we built it
The backend runs on FastAPI with PostgreSQL for storage and Celery (with Redis as the broker) to handle everything that takes time — content generation, image/video rendering, and scheduled publishing — without blocking the app. The frontend is a Next.js and React dashboard styled with TailwindCSS and Radix UI components.
For AI, we built a provider-agnostic LLM layer so we could swap between Gemini, OpenAI, and Claude depending on the task — Gemini's Imagen and Veo models handle image and video generation, while a factory pattern lets us pick the best model for text generation. Brand knowledge lives in Pinecone as vector embeddings, chunked and retrieved at generation time so every post stays grounded in the business's actual documents rather than generic AI filler. Each social platform got its own posting service behind a shared interface, all wired up through proper OAuth 2.0 flows, and the whole thing is fully multi-tenant so every organization's data stays isolated.
Challenges we ran into
Getting five different social platforms to behave through one unified posting interface was harder than expected — each has its own OAuth quirks, media requirements, and rate limits, so we spent a lot of time building a clean abstraction that hid those differences from the rest of the app. Tuning the RAG pipeline was another one: early on, the AI would either ignore the uploaded brand documents entirely or over-rely on them and sound stilted, so we iterated a lot on chunk size and retrieval logic to get natural-sounding, on-brand output. Coordinating async video generation with our Celery task queue without blocking the rest of the pipeline also took a few rounds of rework.
Accomplishments that we're proud of
We're proud that Digilize doesn't just generate content — it actually understands the business it's marketing for, thanks to the RAG grounding. We're also proud of how seamless the multi-platform publishing experience turned out: one piece of content, correctly formatted and posted everywhere, with zero manual resizing or copy-pasting. And we managed to get real ad campaign management and SEO research working alongside the content engine, not as an afterthought — giving small teams a genuinely full-stack marketing tool instead of just another caption generator.
What we learned
We learned just how much invisible work goes into "just post it on social media" — from token refresh flows to platform-specific media specs — and how valuable it is to abstract that complexity away early. We also learned that grounding an AI in real documents changes the entire feel of its output; RAG made the difference between generic marketing copy and something that actually sounds like the business. On the infrastructure side, we came away with a much deeper appreciation for async task design — getting generation, uploads, and publishing to happen reliably in the background taught us a lot about building systems that feel instant even when the underlying work isn't.
What's next for Digilize
Next, we want to add predictive analytics that recommend what to post and when, based on what's historically performed best for each brand. We're also exploring deeper ad automation — letting Digilize not just manage campaigns but actively optimize budget allocation across platforms in real time. Longer term, we'd love to add competitor tracking, so Digilize can benchmark a brand's content and ad performance against others in its space, and expand support to newer platforms as they emerge.
Log in or sign up for Devpost to join the conversation.