The Mission
OODARIS is building the Agentic OS for Retail. We run production agents that plan, decide, and act across pricing, assortment, forecasting, replenishment, and planning. Your mandate is to turn frontier techniques into safe, measurable, repeatable outcomes at enterprise scale. That means multi-tenant data, firm SLAs, deep observability, and safety by design.
What You Will Own
- Agentic workflows end to end. Design and harden multi tool, multi agent flows with planning, retries, guardrails, and contract-enforced inputs and outputs for high stakes tasks.
- Retrieval that does not flake. Lead chunking and indexing, hybrid retrieval, and vector search with clear targets for latency, quality, and cost.
- Inference that scales. Operate a provider-agnostic LLM layer with caching, backpressure, and circuit breaking to deliver sub-second time to first token and low seconds end-to-end.
- Evaluation as a habit. Ship offline evals, CI regressions, red teaming, and online A/B tests that track win rate, grounding, tool success, hallucination rate, and total cost per task.
- Safety and compliance by design. Own input and prompt sanitization, role- and action-level permissions, structured JSON logging with correlation IDs, and audit-ready traces.
- MLOps that keeps up. Manage feature serving, experiment tracking, model registry, rollouts and rollbacks, live dashboards, and clean handoffs to product, data, and customer teams.
- Interoperability built in. Use leading multi-agent frameworks and interop standards so agents collaborate across vendors and clouds.
Impact You Will Drive in 90 Days
- 30 days: Deliver a production-grade RAG slice with automated evaluation dashboards and alerting, plus measurable hallucination reduction on a live retail workflow.
- 60 days: Ship an agent orchestrating two or more tools with explicit recovery paths and schema enforcement, sustaining low latency with caching, backpressure, and circuit breaking.
- 90 days: Reduce cost per task via smart routing and caching, expanding CI to include red team harnesses, parity checks, and rollback playbooks.
Our Environment
You do not need all of this on day one, but you will encounter:
- Agents and interoperability: Leading multi-agent frameworks, interop standards, provider-agnostic routing.
- Cloud and platform: Major cloud providers, containerized services, automated CI/CD pipelines.
- Data and orchestration: In-process analytics, warehouse-native transforms, distributed tasks, event-driven workflows.
- Vectors and retrieval: Postgres with vector search powering multi-modal retrieval.
- Observability and quality: OpenTelemetry-style tracing and metrics, structured logging, evals, guardrails, and orchestrated handoffs.
- Core platform: Asynchronous Python, FastAPI, Redis, Docker, modern IaC, Git-based pipelines.
Education and Fundamentals
- Preferred: BS or MS in Computer Science, Software Engineering, EE, Math, or related field.
- Not required: Equivalent experience is welcome—show mastery of algorithms, data structures, distributed systems, and production engineering through shipped systems, open source, or portfolio work. We hire for ability, judgment, and impact.
What Great Looks Like
- Systems thinker optimizing latency, cost, and correctness simultaneously.
- Fluent with asynchronous Python, end-to-end tracing, and calm incident hygiene.
- Strong product sense—turn messy workflows into crisp agent graphs and contracts.
- Bias to instrument, not guess, and you leave docs that others can use.
Nice-to-Have Superpowers
- Adapter or LoRA experience.
- Retrieval evaluation and reranking expertise.
- Schema mapping across PIM and ERP sources.
- Multi-tool agents with reasoning firewalls, evals, guardrails, and standardized interop.
Why This Role Matters
- Your work changes how merchandisers, planners, and operators make decisions every day.
- Modern stack with typed contracts, pragmatic standards, and a fast but resilient delivery culture.
- High trust and high ownership with real accountability.
How We Work
- Golden Workflow: Plan → Contracts → Code → Tests → Docs → Rollout.
- Typed inputs and outputs everywhere—parity stays green before merge.
- Quality gates: lint + types clean, ≥90% coverage on changed files, happy-path E2E, observability in place before release.
- Security: least-privilege IAM, secrets hygiene, structured audit logs.
- Small, senior, calm team. Pragmatic research. What beats the baseline wins.
The Upside
- Comprehensive health, dental, and vision plus life, disability, and EAP or telemedicine.
- Flexible PTO and remote-friendly culture with Montréal HQ community days.
- Growth path from owning a slice to owning an agent domain to leading a product surface or platform area—growth follows impact.
Hiring Process
- 25-minute intro for mutual fit.
- Deep dive in your strongest area (systems, agents, retrieval, or inference).
- Practical working session with realistic design and problem solving.
- Team loop with product, data science, and engineering.
- Founder conversation for scope and offer.
Ready to Apply?
Apply if you are ready to build multi-agent systems that move the needle. Use the application form on this page or email your profile to careers@oodaris.ai with the role title in the subject line.