Job Description
Senior Software Engineer
\n
Full-Time · Remote (U.S.)
\n
\n
About Velastegui Ventures
\n
We build enterprise AI infrastructure that lets regulated organizations query their unstructured and structured data in natural language. Our platform combines retrieval-augmented generation, federated search, and self-hosted model inference to deliver citation-backed, auditable answers — deployed inside the customer's own cloud boundary. We're growing the engineering team to scale this platform through its first enterprise deployments and into a multi-tenant product.
\n
\n
About the Role
\n
You'll design, build, and operate the backend of our Enterprise Knowledge and Natural Language Query Platform: ingestion, retrieval, inference, and the governance layer that makes AI defensible for regulated customers. This is a hands-on senior role with real architectural ownership — you'll ship on the platform as it stands today while helping design the next generation of it. We're looking for someone comfortable operating anywhere from 5 to 15 years into their career, provided the depth is there.
\n
\n
Where the platform is today, and where it's going
\n
Today, ingestion runs on SQS with Ray-Serve workers, retrieval uses Amazon Bedrock and Bedrock Knowledge Bases inside the customer's AWS boundary, and the platform is deployed via a GitOps supply chain (ArgoCD, Cosign, Kyverno) into customer-owned EKS.
\n
\n
Over the next 12–18 months we're evolving toward a Kafka-based streaming backbone with CDC, self-hosted LLM inference (vLLM / SGLang), self-hosted vector and graph retrieval, and federated search across semantic, lexical, and knowledge-graph modalities. You'll work on both sides of that transition.
\n
\n
The first 6 months
\n
You'll help stand up our first BYOC deployment into an enterprise client on the current SQS + Bedrock stack, own one or two of the six data-source connectors end-to-end, and drive the ingestion pipeline's incremental-sync design.
\n
Expect direct contact with the customer's platform and compliance teams. In parallel, you'll help shape the Kafka-based connector framework and self-hosted inference path that lands on the platform later in the year.
\n
\n
What You'll Do
\n
- Design and operate backend services across an event-driven microservices architecture — ingestion, retrieval, inference, and the connector framework that onboards new data sources.
\n
- Ship on the current SQS + Ray-Serve + Bedrock stack while helping design and land the Kafka + CDC + self-hosted-inference generation of it.
\n
- Extend retrieval quality across semantic, lexical, and (eventually) graph modalities, including fusion and re-ranking.
\n
- Grow the AI serving path: embedding, generation, re-ranking, model routing, and the registries and evals that let us promote models safely.
\n
- Build observability and evaluation into every service you own — tracing, metrics, and quality/drift signals, not just system health.
\n
- Partner with security and compliance on the controls that make SOC 2 and (where applicable) HIPAA and GDPR defensible: encryption, access control, audit lineage, data residency.
\n
- Support enterprise deployments alongside customer platform teams — IaC rollout, cutover, rollback, and the day-two handoff.
\n
\n
Required
\n
- 5+ years building and operating production backend systems, with meaningful ownership of distributed architecture.
\n
- Strong proficiency in at least one of Go, Rust, or Python, and comfort working across a polyglot codebase.
\n
- Solid grounding in distributed systems: asynchronous messaging, event-driven design, and the failure modes that come with them. Direct experience with a streaming or queue technology (Kafka, SQS, RabbitMQ, or similar) and the concepts
\n
behind Change Data Capture.
\n
- Working knowledge of PostgreSQL and at least one non-relational store (vector, document, or graph).
\n
- Fluency with AWS (or equivalent), containers (Docker/Kubernetes), and infrastructure-as-code (Terraform or similar).
\n
- Real interest in — ideally hands-on experience with — applied AI/ML systems: embeddings, vector search, RAG, or LLM serving. This is an AI infrastructure role; curiosity here isn't optional.
\n
- Communication skills to work directly with customers, compliance stakeholders, and product.
\n
\n
Nice to Have (any of these move you up the stack)
\n
- Managed AI services and hosted retrieval (Amazon Bedrock, Bedrock Knowledge Bases, OpenAI / Anthropic APIs) — relevant to the current stack.
\n
- Self-hosted LLM serving (vLLM, SGLang, llama.cpp) or OpenAI-compatible gateways — relevant to where we're going.
\n
- Vector databases and hybrid semantic/lexical retrieval (Qdrant, Weaviate, Pinecone, OpenSearch hybrid).
\n
- Graph or knowledge-graph modeling (Neo4j or similar).
\n
- Large-scale data processing (Ray, Spark, Databricks) and lakehouse patterns.
\n
- OpenTelemetry-based observability and LLM tracing/eval tooling.
\n
- Enterprise IAM (SSO/SAML/OIDC/SCIM) and multi-tenant SaaS.
\n
- Shipping in SOC 2 / HIPAA / GDPR environments, or prior forward-deployed engineering work.
\n
\n
Compensation & Logistics
\n
- Cash compensation: [$210,000 – $250,000] based on experience and location. No equity component.
\n
- Remote within the U.S. Expect an on-call rotation as we support production customers.
\n
- Background check required; the role handles data from customers in regulated industries.
\n
\n
Why Velastegui Ventures
\n
- Ship into a real enterprise deployment in your first year, not a hypothetical roadmap.
\n
- Own architecture on a platform built ground-up for regulated, enterprise-scale AI — not retrofitted from a chatbot demo.
\n
- Small team, direct customer contact, senior autonomy.
\n
\n