I turn expensive, ambiguous data operations into dependable production systems: agents that shorten incident response, platforms that make governed data self-service, and automation that gives senior engineers their time back.
Today I am a Senior Software Engineer in Data & AI Platform Engineering at Red Hat. Across eight years, I have built for regulated enterprise platforms and consumer products serving up to 120M+ users and 500M events per day.
Production MCP and LangChain systems for investigation, release orchestration, metadata intelligence, and operational decision support—with observability and human handoffs designed in.
Self-service data products on Snowflake, Databricks, dbt, Spark, and Delta Lake, with governance, quality, lineage, and cost controls embedded in the paved road.
Cloud-native control planes and developer workflows built with Python, Go, Kubernetes, OpenShift, GitOps, Terraform, and pragmatic automation.
| System | What changed | Scale and stack |
|---|---|---|
| Data Reliability Agent | Automated first-line pipeline triage and reduced initial investigation from ~38 minutes to ~4 minutes. | MCP, LangChain, OpenShift, Langfuse |
| Release Assistant | Orchestrated governed releases and saved 1,000+ lead-engineer hours annually. | Python, GitOps, Snowflake, 150+ compliant data products |
| Self-service Data Mesh | Replaced legacy Redshift/Starburst paths and reduced infrastructure cost by $200K+/year. | OpenShift, dbt, Snowflake, Kubernetes |
| Consumer AI at scale | Built GenAI search, recommendations, and conversational systems for Tata Neu and Beem. | 120M+ users, 500M events/day, 12 Indic languages |
I contribute correctness fixes, security hardening, CI improvements, typing support, tests, and documentation across the data and AI ecosystem.
60 merged upstream pull requests across 28 repositories, including the DuckDB, vLLM, dbt Labs, Red Hat, llm-d, Apache, and LangChain communities.
Latest merged upstream pull requests
- dbt-labs/dbt-jobs-as-code #216 — Fix typos and scope CI permissions ·
2026-08-25 - dbt-labs/dbt-autofix #404 — Fix typos, move dependabot.yml to correct location, fix CODEOWNERS path ·
2026-08-17 - dbt-labs/dbt-autofix #405 — Fix IsADirectoryError crash and bare except clauses ·
2026-08-17 - vllm-project/compressed-tensors #789 — Update Black target-version to py310 ·
2026-08-13 - llm-d/llm-d #2102 — fix: quote variable expansion and remove broken link in guides/ ·
2026-08-10 - duckdb/ducklake #1322 — Replace debug-only assertions with runtime bounds checks in deletion vector parsing ·
2026-08-10
- duckdb/dbt-duckdb #808 — escaped quotes in generated SQL configuration values.
- duckdb/ducklake #1324 — fixed thread-unsafe CRC32 table initialization.
- duckdb/duckdb-rs #822 — returned configuration errors instead of panicking.
- dbt-labs/dbt-codegen #271 — hardened shell variable handling against injection.
Small, public experiments where I explore agent interfaces, developer tooling, data workflows, and useful automation.
| Project | What it explores | Language |
|---|---|---|
| linkedin-mcp-server | LinkedIn automation MCP server wrapping the unofficial linkedin-api | Python |
| snap-a-miro | Convert whiteboard photos into interactive Miro boards using AI vision analysis | JavaScript |
| datadiff | High-performance CLI tool for semantic diffing of tabular data (CSV, Excel, Parquet, JSON) with Git integration | Rust |
| flight-tracker | Local flight price tracker with web UI - supports Amadeus & Skyscanner APIs, daily price monitoring, Indian market optimized | Python |
Recent public activity
- Pushed commits to arijitroy003/llm-compressor ·
2026-08-25 - Pushed commits to arijitroy003/spark ·
2026-08-25 - Pushed commits to arijitroy003/production-stack ·
2026-08-25 - Pushed commits to arijitroy003/llm-d-workload-variant-autoscaler ·
2026-08-25 - Pushed commits to arijitroy003/recipes ·
2026-08-25
Most of my production work ships to Red Hat's private GitLab. This activity graph provides the missing context that a public GitHub contribution graph cannot.
languages Python · Go · Rust · SQL · TypeScript
data Snowflake · dbt · Databricks · Spark · Delta Lake · Kafka · Airflow
ai systems LangChain · MCP · OpenAI · Claude · Mistral · Vector DBs · Langfuse
platform Kubernetes · OpenShift · GitOps · Terraform · Docker · AWS · Azure · GCP
MCA, Jadavpur University · Distributed systems and information retrieval research at ISI Kolkata
Building a serious data platform or production AI system?
Start a conversation
· Explore my work
Project, activity, and upstream contribution data refresh automatically through GitHub Actions.