AI Architect Engineer
Full-time [Location : Hyderabad/Remote] Reports to - Head Solutions
Syren Cloud is building a governed, no-code/low-code enterprise AI platform, and a flagship product powered by it. We're looking for an AI Engineer to own and architect the intelligence layer: the prompts and guardrails that shape agent behavior, and the canonical data model and knowledge graph those agents — and our separate Data Science team's models — reason over.
This role sits closer to architecture than a typical AI engineering position — you're not just implementing against a schema someone hands you, you're deciding what that schema should be.
What you'll do
Architecture & knowledge modeling:
- Own architectural decisions for how intelligence — prompts, retrieval, and agent logic — plugs into our platform's core: where inference happens, how data flows through our workflow orchestration layer, and when a capability belongs in the ontology layer versus a prompt versus the Data Science team's remit.
- Define and evolve the canonical data model — the shared representation of core business entities that every agent, workflow, and downstream model has to agree on, sitting on top of our platform's versioning and governance framework.
- Design automated ontology creation — turning raw source schemas from enterprise systems into structured entities, relationships, and synonyms programmatically, so a new data source onboards without hand-curated mapping every time.
- Build and maintain the knowledge graph connecting the core entities in our domain — the structure our product's retrieval and agent logic reasons over, and the foundation the Data Science team builds its models on top of.
Agent intelligence:
- Design and iterate on agent personas, system prompts, and guardrail logic — role/policy/content-safety checks, autonomy thresholds, and escalation rules.
- Own the evaluation harness: golden tests and prompt-regression checks that catch quality drift before an agent version ships.
- Validate agent behavior across test and production model configurations, working with our internal model registry.
- Tune retrieval quality for knowledge-grounded agents — chunking strategy, relevance testing, and search infrastructure tuning.
How you'll work:
- Build against written specs (spec → plan → tasks) rather than open-ended tickets, and write the specs yourself for the architecture and agent-intelligence surface you own.
- Use AI coding assistants as your primary implementation tool for scaffolding and iteration — your judgment goes into what to build and whether the output is actually correct, not into typing every line by hand.
- Review your own AI-assisted output as rigorously as you'd review a teammate's PR; verification is part of the job, not a step you skip because the code compiled.
- Partner closely with the Data Science team as a consumer of your canonical data model and knowledge graph — you own the structure, they own what gets modeled on top of it.
What we're looking for:
- Experience making architectural calls on a data or AI platform — not just implementing someone else's design.
- Hands-on experience with ontology design, knowledge graphs, or semantic/canonical data modeling — you've built one of these, not just read about them.
- Hands-on experience with LLM-based systems: prompt design, retrieval-augmented generation, and evaluating generative output quality.
- Strong Python fundamentals and comfort working against a modern backend service and its data model.
- Familiarity with agentic or workflow-orchestration patterns — multi-step graphs, tool-calling, trigger → retrieve → act → respond pipelines.
- Comfortable pair-programming with AI coding assistants as a primary tool, balanced with strong code-review instincts.
- Clear technical writing — specs are this role's primary interface with the rest of the team.
Nice to have:
- CPG / Retail / Supply chain domain knowledge
