- 8+ years building and operating production web applications end to end, backend-weighted, with a track record of owning systems rather than tickets.
- Strong Python in production. FastAPI is what we use; Django, Flask or similar is fine if you can move across.
- Solid relational data modelling and migrations against live data.
- Real frontend experience, and the ability to ship usable UI without a design team. Whichever framework you come from, what matters is that you work comfortably on both sides of the line.
- Hands-on experience getting LLM-backed systems into production: agent or graph orchestration, RAG, prompt and context design, and evaluating model output rather than trusting it.
- Fluent daily use of Claude Code or an equivalent agentic coding tool, with a clear view of where it helps and where it doesn't.
- Production experience on a major cloud. We are on AWS: containers, background jobs, secrets, logs, and debugging things that only break in prod.
- Autonomy. You'll get context and a goal, not a decomposed backlog.
- Documentation creation that let someone else pick the work up.
- Professional English
- Connection during LATAM (Chile) Business hours
STACK:
Layer | Technologies |
Backend | Python, FastAPI, SQLAlchemy, PostgreSQL (incl. JSONB), schema migrations |
Frontend | Server-rendered Python templates, lightweight JavaScript, Tailwind CSS |
AI / agents | LangGraph multi-agent pipeline, Gemini and Claude, grounded search (serper.dev), Jina extraction, Pinecone for PDF RAG, vision-model OCR |
Infra | AWS: ECS services and tasks, S3, Secrets Manager, CloudWatch, RDS. Docker, Terraform |
Ways of working | Claude Code as a daily tool, Git branch flow, Jira |
Behind BNamericas is an internal platform that continuously collects data from hundreds of public and private sources, routes it to our research team for verification, and moves it through to publication.
We are rebuilding that platform with AI at its core. You'll own features end to end, from schema to interface, and help turn knowledge held by a few people into a system the whole team can carry. Small team, no ticket factory, and real internal users on what you ship within days.
Product and platform
- Forms and validation: rich, API-integrated data entry for structured content, and the validation rules that keep ittrustworthy.
- Workflow: task assignment, draft to review to publish state machines, and admin tooling that lets the businesschange the rules without a deploy.
- Dashboards and reporting: progress, throughput, and quality visibility for team leads and leadership.
- APIs and integrations: the services connecting this platform to the rest of our systems.
- Auth and access control: roles and permissions across an internal product used by several teams.
- The data layer: schema design, migrations against live data, and keeping queries fast as volume grows.
AI systems
- Multi-agent extraction pipelines: agent graphs that collect, extract and reconcile data from public sources atscale, including retrieval, prompt and context design, model selection, rate limiting and cost.
- Human-in-the-loop review tooling: interfaces where a reviewer sees what the AI proposed, the alternatives itrejected, the evidence behind each, and can override the machine with a justification that gets recorded and kept.
- Audit and provenance: append-only decision trails, so every value can be traced back to a source and to whoeveraccepted it.
- Data-quality gating: the checks between what a model produces and what reaches a customer, so nothing AI originated lands unchecked.
- Source onboarding: new collectors and per-source parsers, the business rules that go with them, and the admin tooling behind them.
At BNamericas we promote inclusive selection processes and are committed to equal opportunities. This job offer is framed within Chilean Law No. 21.015, which encourages the inclusion of people with disabilities in the workforce. If you require any reasonable accommodation to participate in the process, please let us know.
