Soham Pachpande
Soham Pachpande

Hi, I'm Soham.

A human, engineer, and tinkerer. I am currently based out of San Francisco and call the vineyards of Nashik home.
Professionally, I'm a software engineer at Salesforce, where I work on observability and infrastructure.
I'm happiest when I'm walking, reading, biking, or cooking.

About

I grew up tinkering with sensors and scrapers before I knew what "infrastructure" meant: wiring up 66 sensors across my undergrad campus at IIT Gandhinagar to track water and energy use, and separately building a search engine over a few hundred thousand NLP papers, mostly because I wanted to know what I was supposed to be reading. Both turned into publications, which was a nice surprise.

Graduate school at UC San Diego pulled me deeper into distributed systems, and I find designing software systems fun - like a game of legos! At Salesforce: I own the metrics, logging, and application performance monitoring (APM) products that a large chunk of the company's infrastructure and observability runs on, and I've spent the last year or so developing a system to make sense of LLM calls (tokens, tool use, agent traces) and keeping tabs on agentic applications using APM.

Outside of work I'm usually chasing a good cup of coffee, out on a bike ride, or watching a Formula 1 race and getting unreasonably invested in strategy calls I have no control over.

Experience

Salesforce · Senior Member of Technical Staff

May 2023 – Present · San Francisco, CA

I'm a service owner on Monitoring Cloud, the team responsible for infrastructure metrics, logging, and APM agents running across Salesforce's Kubernetes fleet. A lot of my work has been building a control plane that can reconfigure and health-check agents across hundreds of thousands of instances in real time, instead of waiting on a month-long release cycle. More recently I've been building out LLM observability from scratch: designing our APM SDK and pipelines to understand token usage, tool calls, and agentic workflows.

Salesforce · Software Engineering Intern

Jun 2022 – Sep 2022 · San Francisco, CA

Extended our AWS metrics ingestion service to support Alibaba Cloud, designing the cloud-agnostic interfaces that let the same pipeline collect telemetry across providers.

HSBC Technology · Software Engineer, Payments Data Platform

Sep 2020 – Aug 2021 · Remote

Built real-time streaming pipelines on Apache Beam to process SWIFT international payment messages, and batch pipelines to archive them for compliance.

Projects & Publications

SurfStore

A fault-tolerant distributed file storage system in Go, with Raft-replicated metadata and content-addressed blocks served over gRPC.

NLPExplorer

A search and analytics engine over 64K+ NLP research papers, built end-to-end as an undergrad. Published at ECIR 2020.

Water Consumption Monitoring System

Integrated 66 sensors across the IIT Gandhinagar campus to track water, energy, and occupancy in real time. Published at DATA 2020.

Disaggregation for Serverless Computing

Profiled cost and performance trade-offs of a Lambda-based video processing pipeline across disaggregation strategies.

Kubernetes Operator for Telemetry Agents

A hackathon project: an operator that reconciles ConfigMap changes into live telemetry agent lifecycle updates.

Education