Careers
Deploy agents for a living.
Small pods, real systems of record, and a KPI you stay on until it moves.
One pod, one table: the code, the process map and the call with the client's team.
How we work
The client's production, not our backlog.
Remote since 2021
Work from anywhere in México, LATAM or the United States.
Pods embedded in the client
You sit in the client's standups and in their repo.
Evals before code
No agent ships without an eval set that gates it.
You own the KPI
The KPI, the on-call rotation and the cost of your agents.
Open roles
Four roles, one pod.
Forward Deployed Engineer
Remote, México, LATAM or United States · Full time
You own the technical result of one account, against a KPI Contract.
- Python
- TypeScript
- AWS Bedrock AgentCore
- LangGraph
- MCP
- PostgreSQL / pgvector
- Terraform
What you build
- Production agents on Bedrock AgentCore and LangGraph, deployed in the client's AWS account
- Integrations to SAP, Salesforce, HubSpot, Odoo and the client's databases, native or over MCP
- Eval suites that gate every release and keep running weekly once the agent is live
- Approval flows and audit trails the client operates from the Agent Console
- The pilot that reaches production in 8 to 12 weeks, then the next agent on the same account
Process
- 01Application with a short note on a system you took to production and what broke
- 02Technical conversation with a practice lead, 60 minutes
- 03Practical case: design and defend an agent for a real process, async, 3 to 4 hours of your time
- 04Offer, with the range confirmed in writing
Compensation
Range shared in the first conversation.
Deployment Strategist
Remote, México or United States · Full time
You pick the use case, set the baseline, run the KPI Contract.
- KPI Contract
- Discovery Express
- Agent Console
- SQL
- HubSpot
- Google Sheets
What you build
- Use case selection and KPI baselines during the Discovery Express
- The KPI Contract: metric, baseline, target and the commitment gate the client signs
- Business cases the CFO can trace back to the client's own data
- Weekly quality reports and quarterly business reviews with the sponsor
- The expansion plan from the first agent to the next three
Process
- 01Application with one page on a business case you owned end to end
- 02Conversation with a practice lead, 60 minutes
- 03Practical case: turn a discovery transcript into a KPI Contract, async
- 04Offer, with the range confirmed in writing
Compensation
Range shared in the first conversation.
AI Engineer
Remote, México or LATAM · Full time
You build the tools, MCP servers, prompts and evals behind each agent.
- Python
- LangGraph
- MCP
- OpenSearch / pgvector
- LangSmith / Braintrust
- Docker
- AWS
What you build
- MCP servers and tools over the client's APIs and systems of record
- Hybrid RAG pipelines with reranking on OpenSearch and pgvector
- Prompt and routing configurations across Claude, GPT, Gemini, Nova and Llama
- Eval datasets, adversarial tests and red teaming sessions
- Guardrails and PII masking that ship with every agent by default
Process
- 01Application with a link to something you built and a note on how you tested it
- 02Technical conversation with a practice lead, 60 minutes
- 03Practical case: build a tool with its eval set for a given agent, async
- 04Offer, with the range confirmed in writing
Compensation
Range shared in the first conversation.
Voice AI Engineer
Remote, México or LATAM · Full time
You build voice agents and own latency, interruptions and handoff.
- Python
- Parakeet
- Qwen3
- Kokoro
- Vapi / Retell
- Twilio / SIP
- WebRTC
What you build
- Voice agents on the DaCodes stack: Parakeet for speech to text, Qwen3 32B for reasoning, Kokoro for speech
- Telephony over SIP, Twilio and WhatsApp calling, with recording and consent handled per country
- Latency budgets per turn, measured on every release, with barge-in and interruption handling
- Escalation to a person on low sentiment or low confidence, with the transcript attached
- Call evals: transcription accuracy, task completion and hallucination checks
Process
- 01Application with a recording or a repo of a voice system you built
- 02Technical conversation with a practice lead, 60 minutes
- 03Practical case: a voice flow with its latency and eval plan, async
- 04Offer, with the range confirmed in writing
Compensation
Range shared in the first conversation.
Your first month
On a real account by week two.
- 01sandbox · invoice-extraction
- deploy library/invoice-extraction
- deployed to sandbox
- eval run --suite edge-cases
- 3 of 40 cases fail
Week 1: the platform
You deploy a Library agent into a sandbox and break it.
- 02pod · pilot acme-mx
- Forward Deployed Engineerlead
- Deployment StrategistKPI owner
- YouAI Engineer
standup 09:00 · client ops room
Week 2: join a pod
You pair with a Forward Deployed Engineer on a pilot account.
- 03release 1.4 · eval gate
- Accuracy96%
- Policy99%
- PII masking100%
Gate passed→ to productionWeek 3: your first release
Your change passes the eval gate and lands in production.
- 04weekly review · week 4
KPI · first-contact resolution
You presentWeek 4: the review
You present what moved in the client's weekly review.
Demo data
Looking for something else?
Every other DaCodes role is published on dacodes.com.
Here for the platform, not the job?
Two discovery sessions, no cost, and you leave with two or three quick wins whether or not you build with us.
We reply in under one business day.