Agentic Engineer (LLM Agents) — Full-Time
About Vybers
Vybers builds AI-native products that help users get real work done—fast. We’re building agentic systems that can reason over context, use tools, collaborate with humans, and reliably complete tasks end-to-end.
The Role
We’re looking for an Agentic Engineer to design, build, and ship production-grade AI agents. You’ll work across modeling, product, and infra to turn “LLM demos” into robust, observable, safe, and cost-effective agentic workflows.
You’ll own agent behavior end-to-end: orchestration, tool use, memory, evaluation, reliability, and deployment—while partnering tightly with product and design to deliver delightful user outcomes.
What You’ll Do
- Build and improve agentic workflows (planning, tool-calling, reflection/verification, multi-step execution).
- Design tool interfaces and function schemas; integrate with internal/external systems (APIs, databases, docs, web, etc.).
- Implement memory and context management (retrieval, summarization, user/session state, long-term memory patterns).
- Develop evaluation harnesses: offline tests, regression suites, simulation, human-in-the-loop review, and A/B experiments.
- Improve reliability: guardrails, fallback strategies, constraint enforcement, error recovery, and adversarial testing.
- Optimize performance and cost: caching, batching, routing, model selection, latency profiling, token economy.
- Ship to production with observability: traces, structured logs, metrics, prompt/version management, and incident playbooks.
- Collaborate across engineering, product, and research; help set best practices for agent development.
What We’re Looking For
Must-haves
- Strong software engineering fundamentals (design, testing, debugging, performance).
- Experience building LLM-based systems (agents, RAG, tool use, prompt/versioning).
- Comfort working in ambiguous, fast-moving product environments.
- Ability to reason about reliability, safety, and evaluation—not just “making it work once.”
Nice-to-haves
- Production experience with orchestration frameworks (LangGraph, Temporal, Prefect, Airflow, etc.).
- Experience with vector databases / retrieval systems (pgvector, Pinecone, Weaviate, Elasticsearch, etc.).
- Practical applied ML experience (fine-tuning, embedding models, distillation, reward modeling).
- Experience building developer tooling (SDKs, internal platforms, analytics/observability).
- Familiarity with security/privacy patterns for AI products.
Tech Stack
- Backend: Python / TypeScript, REST/gRPC.
- Data: Postgres, Redis, vector search.
- Infra: Docker/K8s, cloud services, CI/CD.
- Observability: OpenTelemetry, metrics/logging/tracing.
- AI: LLM APIs + internal evaluation/experimentation tooling.
How We Evaluate Success
- Agents complete real user tasks reliably with measurable quality.
- Strong regression discipline: improvements don’t break prior behaviors.
- Clear operational ownership: dashboards, alerts, root-cause analysis, iteration loops.
- Product impact: higher task success rate, lower cost/latency, better user trust.
Why Vybers
- High-ownership role at the center of product and engineering.
- Ship quickly, measure outcomes, iterate.
- Work on challenging problems: reliability, tool use, memory, evaluation, and agent UX.
Compensation & Benefits
Competitive salary + equity + benefits. (Add specifics if you want.)
How to Apply
Send your resume + a short note on an agentic system you’ve built or shipped (repo, write-up, or screenshots welcome).
