Srikanth Bellary

• I put agents on telemetry, tickets, runbooks, and the service call graph, then I keep a person on the last step.
• I retrieve from tickets with RAG, and I guide remediation over MCP with the evidence attached.
• The stack I run here is OpenSearch, Lambda, and EventBridge.
• Failure detection sits under six seconds. Manual dashboard watching dropped about 80%.
• Agents I shipped: SRE Recommendation, Remediation Insights MCP, Auto Failover, Correction of Errors, and Confluence.

<6 s

failure detection

80%

less dashboard watching

15 / 20

MCP · REST

SRE platforms · MCP · OpenSearch · Lambda · EventBridge

• I owned the mapping program across 1400+ categories and 30K+ attributes.
• I put confidence scoring and a review queue in front of every proposed map.
• I ran the work on Vertex on GCP.
• Mapping time dropped up to 60%. Attribute-mapping throughput rose 3×.

1400+ / 30K+

categories · attributes

60% / 3×

time · throughput

Vertex · GCP · review queue

• I designed the move from mainframe sources onto Azure business entities.
• I extracted from VSAM and ISAM and kept lineage on the way out.
• I rewrote COBOL data access as SQL, with IDA in the design path.
• I reconciled the landed sets so the target could be trusted.

Azure · VSAM / ISAM · COBOL-to-SQL · IDA

• I led data engineering on RPhAI, the pharmacy-claims ML program for pharmacy workflow.
• I put OCR and NER on the documents that program had to read.
• I ran the pipes across GCP and Azure, with Snowflake as the warehouse and Kafka on the stream.

GCP · Azure · Snowflake · Kafka · OCR / NER

• I built on IHDP — healthcare, financial, clinical, and operational data on one platform.
• I worked the EDI 837 and 835 interchange that the claims traffic actually uses.
• I put graph work on Neo4j and Neptune, and I landed the pipes in Glue.

IHDP · EDI 837 / 835 · Neo4j · Neptune · Glue

• I developed and deployed production Spark ML pipelines in Scala and Python.
• I trained ensembles and tuned them against labeled, feature-engineered sets.
• I put CI/CD around periodic training so the models could be retrained on a schedule.

Spark · Scala · Python · CI/CD

• I worked strategic enablement on the enterprise cloud-migration program.
• I led the big-data track toward a global data lake and off a monolithic core.
• I captured the Capability Maturity Model for the customer and global data platforms.

Global data lake · CMM · cloud migration

• I consulted on the derivatives transformation program as the exchange format moved to ISO 20022.
• I integrated Hadoop with the enterprise applications and the data-layer engine.
• I engineered the pipeline from the production CDH cluster into the central S3 data lake.

Derivatives · ISO 20022 · CDH · S3

• I served as data architect on the Shared Foundation Data program.
• I configured Cloudera Manager for the staging and test clusters.
• I wrote Spark pipelines in Python and Scala.

Cloudera · Spark · Python · Scala

• I led the home-loan data migration onto the mortgage servicing platform.
• I managed 50+ web services consumed by Java, .NET, and mobile clients in the SOA estate.

Home loans · 50+ web services · SOA

• I designed BI reporting around the policy administration platform.
• I implemented the ETL that fed those reports.

BI · ETL · policy admin

• I built warehousing solutions on Kimball methods.
• I designed the star schemas the reports actually queried.

Kimball · star schema · DW

• I developed ETL mappings for the enterprise data systems.
• I wrote the SQL procedures and the transformation logic those mappings called.

ETL · SQL · transformations

• I worked the early ETL on that R&D floor — requirements through delivery.
• I sat with the business stakeholders and turned what they needed into mappings.

ETL · SDLC · Hyderabad