Sahil Kulkarni

Experience

Six-plus years building cloud cost management software, spanning cost optimization, chargeback, vendor cost reporting, and — most recently — agentic interfaces to cost data.

Thomson Reuters

Senior Software Engineer

Nov 2022 — Present

Own and grow the in-house cloud cost management product used by every engineering organization at Thomson Reuters — every tech org carries an OKR to keep cloud waste under 5%, tracked through this platform. Have identified $30M+ in annual savings across an expanding set of waste categories, up from $8M when the product was younger. Provide technical guidance and feedback to a team of 5–6 engineers. Advise on Azure and OCI cost design decisions without owning hands-on engineering on those clouds; AWS and GCP are the primary focus.

Product-to-application cost mapping

Built an algorithm that parses complex app-to-product relationships and overlays revenue data, enabling true unit economics — cost and revenue per product — for C-level decision-making.

FinOps chatbot agent (Bedrock AgentCore)

An MCP server extends the in-house cost platform’s APIs so LLMs and agents can query cost data directly. Evaluated a Lambda + API Gateway POC (rejected on latency) and an internal EKS-hosting platform (rejected — CI/CD too weak for the pace of AI iteration) before waiting for Bedrock AgentCore’s GA, POC’ing it, and getting stakeholder and architect sign-off. Live in production since July 2026 with 100+ internal users, open to all employees.

GCP daily resource-level chargeback

Extended chargeback beyond AWS to daily, resource-level granularity on GCP — a major recent focus area as the platform’s scope has expanded to multi-cloud.

Chargeback re-architecture on CUR v2 + Athena

The original chargeback pipeline called the Cost Explorer API monthly; as adoption grew, requests for more dimensions (region, platform, cost allocation tags) made API-call volume unsustainable. POC’d the newly released Cost and Usage Report v2 and rebuilt the pipeline on Athena queries — extensible to new columns without exponential API growth, and immediately reusable to onboard cost data for new M&A organizations.

Multi-vendor cost data normalization pipeline

Defined a normalized ingestion format and worked with platform teams on what to send and how; built the ingestion, validation, and transformation layer that turns raw vendor exports into reporting-ready data across Datadog, Snowflake, Anthropic, OpenAI, and others.

Tokenomics dashboard

Solo-designed and built a dashboard normalizing AI cost and usage data across AWS, Azure, OCI, GCP, and vendors including Anthropic, OpenAI, Claude, Cursor, and LiteLLM.

Cloud modernization scorecard

A side project generating reports and recommendations for IaaS-to-PaaS modernization, quantifying the dollar savings of each move. Previously Thomson Reuters had no data-driven view of which applications to prioritize; this scorecard gave that prioritization a dollar figure.

Capgemini

Analyst → Senior Analyst → Associate Consultant → Consultant

Aug 2018 — Nov 2022

Built the same cloud cost management product from scratch for Thomson Reuters, Capgemini’s client, progressing through four titles over four years. Identified roughly $8M in annual cloud savings as the platform matured. Moved to Thomson Reuters directly to continue leading the product as it scaled into the multi-service platform it is today.