# Bytery

> Bytery is a small AI engineering studio. We build AI agents, data pipelines and provision the infrastructure to deploy and scale in production.

Founded 2013 by Roberto Galán. Based in Argentina, working with teams in the US and Europe.

## Services

- Agentic workflows: Multi-step agents that plan, call tools and write back to your systems (ERP, CRM, ad platforms). Tool calling, MCP, evals and guardrails, so you know when they regress.

- Document & order intake: Voice notes, photos, PDFs and spreadsheets turned into structured records. Every line gets a confidence score; only the uncertain ones reach a person.

- Generative media pipelines: Script, voice, music, captions and edit, assembled automatically with vision checks. Your editors review instead of assemble.

- Internal AI tools: Chat over your data, copilots, back-office assistants. Hosted models or self-hosted open weights, whichever your security team signs off on.

- Data pipelines: ETL, scraping at scale, embeddings and retrieval, BI. The plumbing that keeps every AI feature working at 2am.

- Full-stack product: Web and API products from schema to ship.

## Selected work

[Watch the portfolio videos](https://robertogalan.com/portfolio/)

- WhatsApp-native AI order desk: Distribution & ERP. Orders arrive as voice, text, photos, XLSX and PDF; an agent parses and matches every line before one-click SAP export. 99.8% of sales orders confirmed automatically.

- Autonomous video ad pipeline: Media operations. Multi-model pipeline (Claude, Gemini, Suno, WhisperX, FFmpeg) that writes, voices, scores and edits ads. 30 h saved every week.

- Creative analysis engine: Meta Marketing API data joined with vision-model reads of each creative, gated by evals. 120 analyst hours saved weekly.

- Retrieval-grounded concept generator: Ad concepts grounded in 13k+ enriched customer profiles. 10x faster ad production.

- LLM-assisted audit back office: Regulated finance. 30 h saved per audit, 16 h saved on weekly reporting.

- Ticketing & venue platform: Hospitality, Las Vegas. In-house e-commerce and ticketing that saved millions in third-party fees.

## How we work

- Your stack or ours: we work inside your repos and infrastructure, or run it on our own infra and architecture. Whichever fits.
- Ship: first production release in about a week.
- Hand off: 100% code ownership transferred. Zero frameworks we evangelize.
- Measure: every system ships with evals and a number the business already tracks (hours saved, orders confirmed, fees avoided).

## Stack

Model- and vendor-agnostic. Preferred:

- Models: Claude, Gemini, OpenAI, open-source models on edge devices
- Agents: tool calling, MCP, evals & guardrails, RAG, custom orchestration harnesses
- Media: Seedance, Nano Banana, Suno, ElevenLabs, WhisperX, OpenCV, FFmpeg
- Backend & data: Python, Flask, PostgreSQL, SQLAlchemy, Pandas, vector DBs, GraphQL
- Infra: AWS, Azure, Docker, Nginx, Linux, CI/CD, observability

## Contact

- [Email](mailto:hi@bytery.com): A human replies, usually within the business day.
- [Portfolio videos](https://robertogalan.com/portfolio/): One-minute videos of the systems above.
- [Founder](https://robertogalan.com): Roberto Galán, resume and full track record.

## AX

- [llms.txt](https://bytery.com/llms.txt)