Please mention DailyRemote when applying
See how much of this job your resume covers, and what’s missing.
Want a recruiter to go through it line by line?
Get professional reviewQuestions interviewers often ask for this role, with sample answers.
One click makes a copy of your resume for this role, scores it against the posting, and proposes edits line by line. Your default resume is never changed.
Upload your resume and we draft a letter for this exact role, tailored to what it asks for.
You will own the application architecture for NovaOS, designing reusable patterns for agentic workflows, knowledge systems, and enterprise-grade services. You will act as a key technical leader, partnering with platform leadership to ensure production readiness and consistent engineering standards across all customer solutions.
20 Nov 2026 · @Ashu
OrbitronAI builds NovaOS, a production agentic operating system deployed inside regulated enterprises in energy, oil and gas, and aviation. NovaOS already has dedicated platform leadership. This role owns what gets built on top of it: ComplyNova, FlowNova, AssetNova, Contract Intelligence, and customer-specific solutions.
You make those applications consistent, secure, multi-tenant, and shippable. You are the right hand to the Head of Platform and the CTO's go-to architect for application design, customer solution architecture, and production readiness.
This is a hands-on role. You design, review, and help ship real architecture across full-stack services, agentic workflows, knowledge systems, evals, and deployment. We need equal depth in enterprise product engineering and production AI. Strength in only one will not be enough.
These are the live problems you would walk into. Each one needs an architect who has shipped it, not studied it.
One architecture across products. Four products and a growing set of customer solutions share the same needs: frontend-to-agent flows, workflow execution, audit trails, human approval. Without common patterns, each team invents its own. NovaOS solutions must stay generic, never built for a single product.
Long-running agents. We are moving from assistants to agents that pursue goals over weeks or months. That demands durable execution, resumable state, approval gates, cost control, and clear ownership of what an agent may do unattended.
Enterprise knowledge, beyond plain RAG. Customers need documents turned into knowledge their teams and agents can trust. We are deciding where ontologies, knowledge graphs, GraphRAG, and wiki-style knowledge stores beat vector retrieval, and how to keep them simple to maintain and always current.
Memory that is safe and useful. Agents need short-term, long-term, per-user, per-tenant, and per-case memory. What gets remembered, what gets forgotten, and who can see it are architecture decisions, not prompt tricks.
Proving agents are good. Enterprise buyers ask hard questions on hallucination, data governance, and security. We need eval-gated releases, grounded citation checks, regression suites, and production telemetry that feeds back into quality.
Connecting to unreliable customer systems. ERPs, CRMs, and document stores fail in creative ways. One connector declaration should serve both REST and MCP, enforce RBAC at the dataset level, and carry provenance from raw data to knowledge.
Deploying where the customer requires. Cloud, BYOC, on-prem, and sovereign environments, sometimes with on-prem models, plus the security reviews and compliance flow-down that come with regulated customers.
Application architecture on NovaOS
Reusable patterns for frontend-to-agent flows, backend-for-frontend, domain services, agent orchestration, workflow execution, audit trails, and human-in-the-loop review.
Reference implementations and engineering standards, so no product team builds its own stack.
The mapping from application needs to NovaOS capabilities (execution gateway, LLM gateway, MCP gateway, identity and RBAC, memory, evaluation, observability, deployment), and the gaps you turn into clear platform requirements with the Head of Platform.
Production readiness reviews and major design decisions for products and customer solutions.
Full-stack product architecture
Enterprise services in Java/Spring Boot (Go where it fits), React/TypeScript, PostgreSQL, Redis, Kafka, and cloud-native infrastructure.
API design, service boundaries, domain models, and data flows.
A concrete definition of what 10x growth means for NovaOS applications, and the path there.
Pragmatic build-vs-buy calls across services, AI workflows, and integration layers.
Agentic, knowledge, and memory architecture
Multi-agent and long-running workflows: planner-executor patterns, durable execution, tool access, approval gates.
The knowledge layer: ingestion, chunking, embeddings, ontologies, knowledge graphs, retrieval quality, citations, access control, and tenant isolation.
The memory system: scopes, retention, forgetting, and permissions across user, tenant, and case.
Governed agent-to-tool access through MCP: permission boundaries, auditability, versioning, and discovery.
LLM use in production: model selection, fine-tuning versus prompting versus retrieval, latency, cost, fallback, and safety, including prompt injection and PII handling.
Evaluation-driven releases: test suites, quality gates, shadow testing, regression and hallucination checks, and telemetry feedback loops.
Integration and deployment
Reliable integrations with ERPs, CRMs, document stores, enterprise data platforms, and industrial systems, using retries, queues, idempotency, and reconciliation.
Application readiness for cloud, BYOC, and on-prem, with RBAC, data segregation, auditability, and customer security reviews.
Technical leadership
Architecture reviews, mentoring of staff and senior engineers, and standards that teams actually follow.
Turning ambiguous customer and product asks into executable technical direction.
12+ years in software engineering, with at least 4 years as ex tech Founder, Principal, Staff, Architect, or senior technical lead at a product or platform company.
Strong full-stack architecture: Java/Spring Boot, React/TypeScript, PostgreSQL, Redis, Kafka, and production-grade API and service design. You can review, influence, and shape Spring Boot services.
Enterprise SaaS or platform applications designed, shipped, and operated for real customers.
Production AI and agentic systems, not demos: RAG pipelines, LLM-integrated applications, multi-agent workflows, tool use, memory, evaluation, and safety patterns.
Experience with durable orchestration or workflow engines such as Temporal, LangGraph, or an equivalent.
Strong enterprise integration experience with APIs, CRMs, ERPs, document systems, and unreliable external systems.
Deep understanding of RBAC, multi-tenancy, data segregation, auditability, observability, and enterprise security.
Systems where latency, cost, scale, reliability, and correctness were explicit design constraints.
Building on a platform without duplicating platform concerns, and partnering with platform leadership instead of competing with it.
Clear communication of trade-offs to engineers, product, founders, and enterprise customers.
Startup DNA and an AI-native working style: you have thrived in a high-velocity startup, you make progress without hand-holding, and you use AI tools daily for design, code, and review. You read AI-written documents quickly and critically, and you challenge them rather than skim or dismiss them. You are comfortable in ambiguity and have done 0-to-1 before.
The first five are where our hardest problems sit today. Shipped experience in several of them will set you apart.
Ontologies and knowledge graphs. You have modelled an enterprise domain as an ontology or graph, shipped it to users, and kept it maintainable. You know when GraphRAG beats plain RAG and when it does not.
Memory systems for agents. You have designed memory with explicit scopes, retention, forgetting, and access control, and measured whether it improved outcomes.
Evals. You have built eval suites and release gates for LLM or agent systems: golden sets, regression checks, hallucination and citation grounding checks, LLM-as-judge with calibration, and production feedback loops.
Model fine-tuning. You have fine-tuned or distilled models for a real product, including data curation, evaluation against a strong prompted baseline, and the cost, latency, or on-prem reasons that justified it. You can say when not to fine-tune.
Document intelligence. Extraction, compliance workflows, and enterprise knowledge systems where citations and access control are non-negotiable.
Experience at an AI-native, agentic, enterprise AI, or developer platform company.
MCP or similar agent-to-tool integration patterns, with governance.
Regulated enterprise customers in energy, oil and gas, aviation, finance, government, or critical infrastructure.
On-prem, BYOC, hybrid, or sovereign deployments, and applications that passed security reviews, VAPT, and compliance audits.
Production systems on GCP (GKE, Cloud Run, Pub/Sub, BigQuery, IAM, VPCs), and Kubernetes-based enterprise deployment with Helm, Terraform, GitOps, and network policies.
Prompt injection defense, grounding, and safety patterns in agentic contexts.
You will succeed if you are a hands-on product architect who can discuss Spring Boot service boundaries in one meeting, ontology design in the next, retrieval quality and eval gates after that, and a customer's on-prem constraints after that, without losing sight of shipping speed. You have strong opinions and low ego. Your job is to make NovaOS more useful, make applications more consistent, and give the CTO confidence that products built on it are technically sound. You are here to help teams ship better systems faster, not to write architecture documents nobody uses.
This role is not for:
Someone who wants to govern from a distance.
Someone who has only built AI demos, prototypes, or internal experiments.
Someone strong in traditional SaaS without deep production agentic experience, or strong in AI workflows without enterprise full-stack design depth.
Someone whose knowledge graph, evals, or fine-tuning work stayed in research and never reached users.
Pure consulting profiles without long-term product or platform ownership.
Someone who has not worked in a high-velocity startup, and needs structure, detailed direction, or hand-holding to make progress.
Someone who has not upskilled in AI tooling, or does not use AI daily in their own architecture and engineering work.
Someone who is not used to reading, reviewing, and challenging AI-written documents, specs, and designs, which are a large share of how our teams communicate.
Someone who cannot operate in ambiguity, take ownership, and drive clarity across teams.
First 30 days
You understand NovaOS capabilities, the current application architecture, and active customer priorities.
You have a working rhythm with the Head of Platform, the CTO, and engineering managers.
You have reviewed the key applications and named the top architectural risks, duplication, and gaps.
You have drafted reference patterns for frontend-to-agent flows, backend services, knowledge and memory usage, MCP tool access, and eval gates.
You have unblocked at least one major application or customer technical decision.
First 60 days
Application patterns are standardized across the major NovaOS-based products, and duplicated architecture is shrinking.
A written decision exists on the knowledge layer: where ontologies and knowledge graphs are used, where plain retrieval is enough, and how it stays maintainable.
A memory model with scopes, retention, and access rules is agreed with the Head of Platform.
At least one product ships behind an eval gate, with production readiness criteria covering observability, safety checks, tenant isolation, and deployment.
Recurring application needs are feeding the platform roadmap.
Senior engineers are measurably stronger through design reviews, mentoring, and reference implementations, and the CTO treats you as the trusted authority on NovaOS application architecture.
Expect deep technical discussion on:
Full-stack service decomposition, Python, Go, Java/Spring Boot API design, and React/TypeScript frontend architecture for enterprise apps.
Multi-tenancy, RBAC, data segregation, and auditability.
RAG design, retrieval quality, grounding, and citations, plus ontologies, knowledge graphs, and when each earns its cost.
Agent memory design and long-running, durable multi-agent workflows.
Evals and release gates for agentic systems, and fine-tuning versus prompting versus retrieval.
LLM production concerns: latency, cost, safety, fallback, and observability.
MCP and tool governance, and integration with unreliable customer systems.
Cloud, BYOC, on-prem, and hybrid deployment constraints.
Trade-offs in a fast-moving startup, and how you partner with platform leadership, product, and the CTO.
Logistics
BE, MS, or PhD in Computer Science, or equivalent hands-on experience.
Remote work possible for eligible candidates.
Stop the endless job search. Our AI finds and applies to the best jobs for you.
Featuring 217,732+ Jobs in Architect
Answer easy questions
217,732+ jobs across 15+ categories
Get your best job matches
Only hand-screened, legit jobs
Find a remote job faster
No ads, scams, or junk
“I was the first applicant for a remote marketing position that got listed on the company website the same day I applied. Had an interview within 48 hours!”