Romit S.

I am a hands-on Data & AI Platforms Architect and Lead Systems Engineer with 14+ years of experience designing, building, and optimizing distributed systems from scratch. I specialize in building petabyte-scale, near-real-time streaming infrastructure and production-grade, self-hosted AI platforms across AWS, GCP, Azure, and on-premise environments. Rather than relying on generic wrappers, I engineer low-latency, scalable backend architectures using high-performance languages like Golang and Rust, alongside Python for advanced AI orchestration. Advanced AI & MLOps Infrastructure • Self-Hosted AI Platforms: Deploying open-source LLMs at scale using Kserve on Kubernetes, vLLM, and Ollama. • Agentic AI & Orchestration: Building complex context-aware applications using LangGraph, LangChain, and MCP (Model Context Protocol) Servers. • Advanced RAG & Semantic Data: Engineering hybrid semantic mapping layers that merge unstructured and structured data using heuristics, vector storage, and memory systems. • MLOps Frameworks: Setting up production pipelines with Kubeflow, MLflow, and LLM observability via Langfuse. • Modernization: Re-architecting and transitioning legacy data platforms into high-performance, AI-first ecosystems. Deep Data Engineering & Core Systems • Data Internals & Optimization: An open-source contributor to core data products (like Airbyte). Passionate about database internals, indexing, and high-performance tuning. • Streaming & Batch Infrastructure: Architectural expertise in building resilient pipelines using Kafka, Flink, and Spark, complete with CDC, data observability, and proactive alerting. Domain & Data Types Handled I have designed systems specifically optimized for: • Telemetry, system logs, & clickstream data • High-throughput payment and eCommerce data • Low-latency video and sports data analytics • Affiliate marketing attribution systems Core Technology Stack • AI Ecosystem: LangGraph, LangChain, Kserve, vLLM, Qdrant, Langfuse, Ollama, MCP • Big Data & Streaming: Spark, Kafka, Flink, Presto, Dremio, Apache Hudi, Delta Lake • Warehouses & OLAP: Snowflake, Druid, ClickHouse, Redshift, SingleStore (MemSQL), QuestDB • Databases: Postgres, MySQL, Cassandra, DynamoDB, DuckDB • Languages: Golang, Rust, Python, Scala, Java • Orchestration & Data Ops: Airbyte, Fivetran, Dagster, Airflow, Apache NiFi, Kubeflow, Elasticsearch, OpenSearch • Platforms & Cloud: Databricks, Snowflake, Cloudera, Supabase, Aiven | AWS, GCP, Azure • Infrastructure: Kubernetes, Docker