AI Research Engineer, Agentic Post-Training at Tether

Confirmed open

Improve tool use, reasoning and multi-step behaviour through agent-focused model post-training in Hanoi.

Tether

Location: Remote
Advertised region: Hanoi, Vietnam
Employment: fulltime_permanent
Work arrangement: remote

Create a free application profile — Keep your skills and projects in a reusable profile. Private by default; you choose when to share. Apply for this job directly with the employer.

Role overview

This research role improves pretrained models so they can plan, use tools and complete multi-step tasks, including on resource-limited devices. Work ranges from agent training data to large-scale post-training and evaluation. The employer lists Hanoi, Vietnam.

Responsibilities

Research post-training methods for agent behaviour, tool calling and reliable multi-step reasoning.
Improve factuality, instruction following, function use and multi-agent coordination.
Build data pipelines, training workflows, benchmarks and evaluation systems for agent models.
Curate tool-use trajectories and interaction data, diagnose failures and improve baseline model quality.

Qualifications

Computer science or machine-learning degree; an advanced degree and strong AI publications are preferred.
Experience with multimodal post-training data and workflows for agents and tool use.
Distributed training experience on multi-node GPU systems and measurable improvements in reasoning or tool use.
Relevant open-source code, datasets or models and a research publication record.

Technology in this role

CAN: This research role improves pretrained models so they can plan, use tools and complete multi-step tasks, including on resource-limited .

Role-specific candidate guide

Preparation notes based on the employer’s stated responsibilities and qualifications.

Understanding the scope. This AI Research Engineer (Agentic Post-training) - 100% Remote Worldwide position at Tether is best assessed through the work described in the employer’s posting, rather than the title alone. The available role overview says: “This research role improves pretrained models so they can plan, use tools and complete multi-step tasks, including on resource-limited devices. Work ranges from agent training data to large-scale post-training and…” Read that statement together with the responsibilities below to understand the likely balance between delivery, coordination and technical decision-making. The role belongs to the AI and data area, but that label is only a guide to the work; the employer’s own requirements remain decisive. Before applying, check the original listing for reporting lines, product or team assignment, employment terms and any details that may have changed since this copy was checked. Where this page says a fact is not specified, treat it as an open question rather than an implied promise.

What the responsibilities signal. The posting specifically names “Research post-training methods for agent behaviour, tool calling and reliable multi-step reasoning.”; “Improve factuality, instruction following, function use and multi-agent coordination.”; “Build data pipelines, training workflows, benchmarks and evaluation systems for agent models.”. For each duty, identify a project where you performed comparable work and be ready to describe the context, your own contribution, constraints, decisions, and measurable result. Explain how you moved from the initial need to implementation, review, release or operational follow-through, where those stages apply. If the work involved other teams or users, distinguish your decisions from the group’s contribution. Avoid presenting a technology keyword as proof by itself: connect it to the problem you solved, the scale or complexity involved, and what changed after your work. The source may contain additional duties; review it directly before preparing your application.

How to read the requirements. The listed qualifications include “Computer science or machine-learning degree; an advanced degree and strong AI publications are preferred.”; “Experience with multimodal post-training data and workflows for agents and tool use.”; “Distributed training experience on multi-node GPU systems and measurable improvements in reasoning or tool use.”. Mark each requirement as demonstrated, adjacent experience or a genuine gap, then use a concise example for the first two categories. A strong example explains what you personally owned, how you made trade-offs, how you validated the result and what you learned. If the posting specifies years of experience, language ability, certifications, location or eligibility, address those facts plainly in your CV or application rather than relying on the employer to infer them. Do not claim a qualification from exposure alone. For a requirement you do not meet, describe relevant transferable experience accurately and ask the recruiter whether it is mandatory.

Technology and working evidence. The technology terms associated with this listing are CAN: This research role improves pretrained models so they can plan, use tools and complete multi-step tasks, including on resource-limited .. Treat this as a set of posting-derived signals, not a guarantee that every tool is used daily or that the list is complete. For each term relevant to your experience, prepare one concrete example: the system or workflow, the part you owned, the alternatives considered, and how you tested reliability, security, performance or usability as appropriate. If you have not used a named tool, explain the closest related system you have worked with and how you would ramp up; do not relabel that experience as direct production use. If the source describes technologies beyond this list, prioritize its exact wording.

Evidence to put in the application. Tailor the first half of your CV to the named work instead of using a generic summary. Lead with two or three outcomes that match the responsibilities, and include enough context to show complexity, scope and your role. Add technical detail where it substantiates an outcome: architecture, interfaces, data or traffic scale, test strategy, deployment approach, incident response or collaboration, as relevant to your real projects. Keep confidential employer information private; describe.

About Tether

Tether develops digital-token products, including USDT, alongside projects in peer-to-peer communications and AI. The employer describes a globally distributed remote team.

Employer website

How to apply

Review the original posting and apply through the employer’s careers page.

View source and apply Apply directly

Source checked: 2026-10-17