Embedded Infrastructure Engineer, Chanakya

Sarvam deploys its AI platform inside client environments rather than only as a hosted service, and Embedded Infrastructure Engineers build the data infrastructure those on site deployments run on. You would work alongside embedded data scientists and strategic deployment engineers, making sure terabyte scale datasets can be ingested, stored, queried and served to AI reasoning engines reliably. The data is genuinely varied, spanning structured records, documents, imagery, audio and geospatial, and the volumes are described as terabyte scale persistent stores with sustained large daily ingestion. You would make the calls on storage architecture, indexing strategy, pipeline orchestration and performance, and the posting is clear that this often happens in constrained, air gapped or operationally sensitive environments where managed cloud services and standard enterprise tooling are simply not available. You own the reliability and performance of the infrastructure layer for the accounts assigned to you, which makes the embedded part of the title literal rather than decorative.

What the posting requires

- 4 to 8 years in data infrastructure, data engineering, platform engineering or site reliability engineering, ideally somewhere operating at significant data scale
- Direct experience managing multi terabyte data stores, having personally built or operated systems holding 10TB or more with sustained high throughput ingestion
- Deep working knowledge of at least two of PostgreSQL, MongoDB, Elasticsearch, ClickHouse or comparable systems, including tuning, indexing, partitioning and operational management
- Production ingestion pipelines built with Kafka, Spark, Airflow, Flink, dbt or equivalent
- Strong Python and/or Go, writing production infrastructure tooling and automation
- Solid understanding of storage systems and formats: object storage such as S3 or MinIO, columnar formats such as Parquet and ORC, and how to pick the right layer for a workload
- Containerisation and orchestration with Docker and Kubernetes in production
- Infrastructure as code and deployment automation with Terraform or similar

Counts as a bonus, not a requirement

- Vector databases or embedding stores such as Milvus, Weaviate, Qdrant or pgvector
- Deploying and operating infrastructure in air gapped, on premise or hybrid environments

What you would actually be doing

- Designing and operating storage architectures across relational, document, vector and object stores at terabyte scale and across multiple data types
- Building ingestion pipelines that reliably handle daily influx across both batch and streaming workloads
- Making the decisions on indexing strategy, pipeline orchestration and system performance
- Owning reliability and performance of the infrastructure layer for your assigned accounts

Location and working pattern

Delhi based, working closely with teams deployed at client sites.

Worth knowing before you self reject

As with its other openings, Sarvam notes it wants people who can own the outcomes described rather than people matching every line of the specification.

A good fit if

You are a data or platform engineer comfortable working at customer sites and making architecture decisions without a managed cloud safety net.

Think twice if

You want to stay on internal systems with modern tooling. Air gapped and operationally sensitive environments are the norm here, not the exception.