Mohan Ranganath Varma Vummadisetty

Problem — Legacy MySQL queries and stored procedures needed Snowflake equivalents as part of a wider migration led by senior engineers.

Work — Converted queries and stored procedures to Snowflake SQL, resolved dialect incompatibilities, and validated the converted logic against the originals. Scope was the code conversion — senior engineers handled the data migration itself.

Outcome — Converted workloads validated against source behaviour. This was the introduction to Snowflake that everything since is built on.

Problem — An interoperability proof-of-concept with hard client limits: process the full workload in under 45 minutes for under $4 of compute, writing out to Apache Iceberg on S3-compatible storage.

Work — Redesigned and iteratively tuned the workload, then got under the cost line by moving the account from Enterprise to Standard edition — same compute, lower credit rate, so the trade is enterprise features rather than performance. Isolated an Iceberg write failure to bucket permissions with a controlled cross-provider test.

Outcome — Processed 1 TB to Iceberg inside both limits, and drove the write failure to resolution with Snowflake support and vendor engineering. Proof that performance engineering is a feature, not an afterthought.

Problem — A private-equity roll-up had inherited on-premises SQL Server through its acquisitions, including procedural execution chains scheduled by clock time with no real dependency graph behind them.

Work — Migrated 150+ tables and 50 stored procedures — over 1 TB — extracting and loading in parallel via Spark with stage-wise reconciliation of row counts and value ranges. Mapped the real dependency graph and rebuilt it as a 10-to-14-level Snowflake Task Graph.

Outcome — The batch window fell from 7–8 hours to roughly 2, with dependency-aware orchestration replacing clock-time scheduling. Automation moved the data; the procedural logic was translated by hand.

Problem — PostgreSQL reporting queries were correct but slow, and had to move to Snowflake without changing results — or growing the warehouse to get there.

Work — Converted 35 PostgreSQL reporting queries to Snowflake SQL and refactored shared CTE logic into clustered Dynamic Tables chained by target lag. Delivered semantic models and a Snowflake Intelligence agent through UAT and go-live.

Outcome — Runtimes fell from over 13 minutes to under a minute — several of them sub-second — with the warehouse size held at S. Same data, a fraction of the wait and none of the extra spend.

Problem — A commercial analytics platform for a NASDAQ-listed life-sciences company needs continuously evolving datasets for patient journey, referral, forecasting and marketing analytics — under strict data-quality expectations.

Work — Own a Snowflake enterprise data warehouse integrating eight data sources behind master data management and semantic layers. Built patient journey, referral and line-of-therapy models from claims — inferring each referring physician from earlier claims to map referrals into the client's treating radiologists. Designed a competitive intelligence pipeline that runs entirely inside Snowflake: a weekly task chain searches news on nine competitors, crawls through a self-rebuilding egress allow-list, and has Cortex classify and score each article, with content hashing so unchanged articles are never paid for twice. Geocoded clinical trial sites with a tiered address fallback, raising mappable rows from 27.6% to 99.7%, validated against 5,738 known addresses at a 1.35 km median error.

Outcome — The client credits the weekly outputs with field-sales leads that led to treatments for patients who would not otherwise have been reached. Underneath it: gated weekly orchestration where a failed run reruns only its failed steps, an SFTP feed and a Microsoft Graph API stored procedure replacing a costly managed ingestion service, Delta Sharing from Databricks with OIDC federation, and 22 Streamlit apps that let the client update and control data through a UI instead of code — forecast data entry among them, in production use.

Problem — Two engagements running concurrently with Delcath needed delivery accelerated: an SAP-to-Snowflake conversion, and a marketing-data ingestion build for an Amazon/marketplace advertising agency.

Work — Hand-translated SAP ABAP procedural logic into set-based Snowflake SQL at Sattva — the partner accelerator migrates the data but cannot convert the code — and built Funnel.io ingestion for advertising data at Blue Wheel Media. Partial scope on both: specific pieces, not whole projects.

Outcome — Together with the other engagements these complete a four-source-platform conversion record — SAP ABAP, T-SQL, MySQL and PostgreSQL to Snowflake SQL. Almost no Snowflake engineers carry the SAP side of that.