Gaurav Jain

Principal Engineer

Cortex — CoWork / Snowflake Intelligence

• Built zero-latency secure sandboxes for agent code execution — model-generated code starts instantly, fully isolated, without leaving the customer's data boundary.
• Designed strong access controls enabling fine-grained database access policies, giving customers precise control over the scope of agent capabilities.

Snowpark

• Scaled Snowpark from 3% of product revenue to a $100M+ run rate — over 100 million queries per day from more than half of Snowflake's customers.
• Rebuilt sandbox infrastructure on top of gVisor, strengthening isolation and program compatibility at zero performance cost — 1.5% faster than the legacy in-house sandbox on full benchmarks.
• Optimized the gVisor runtime — cut virtual-memory bookkeeping (VMA entries) by 182x on memory-heavy workloads, and improved ARM compatibility for ML libraries with better CPU virtualization.
• Built a secure pipeline for Python package delivery with sub-second dependency solves — a 99.95% solver-cache hit rate that cuts query initialization latency by ~85% in production.
• Publications:
  • SEE++: Evolving Snowpark Execution Environment for Modern Workloads
  • Snowpark: Performant, Secure, User-Friendly Data Engineering and AI/ML Next To Your Data

Tech Lead / Engineering Manager (TLM)

TensorFlow

• Delivered core functionality for the next-generation TensorFlow 2.0 framework, including development of GPU/TPU accelerator management, missing kernel functionality, as well as API testing and consistency.
• Optimized the core eager runtime, achieving a 2x training speedup for dynamic models, matching compiled graph mode performance on ResNet50 and improving TPU startup time by 50x.

ML for Systems

• Built and evaluated applied-ML techniques that replaced hand-tuned heuristics in Google's backend systems.
• Projects included ML-guided query planning — Kepler, a 2.41x geometric-mean speedup over PostgreSQL's optimizer with near-zero regressions — ML-optimized block caching, and ML-tuned compiler optimization passes.
  • Published: Kepler: Robust Learning for Faster Parametric Query Optimization
  • Patented: Autonomous Column Selection for Columnar Cache

Founding Engineer / Engineering Manager

• Early engineer as Rubrik grew from 10 people to more than 1,000 within 4 years.
• Patented work across the distributed file system and data management stack:
  • Cluster-Based Network File Server — a highly available file system optimized for ingesting backups from virtualized environments.
  • Throttling Network Bandwidth Using Per-Node Network Interfaces
  • Chunk Allocation

Polaris

• Led backend development for Rubrik's ML-powered cloud data management platform.
• Designed key components and service interactions, such as RPCs and database schema management.
• Built remote communication protocol with transport fault tolerance and agent discoverability.

Jarvis

• Implemented cross-language performance & monitoring framework for comprehensive system analysis.
• Enabled proactive customer support with real-time log & stats alerting engine using AWS stack.

Forge

• Built cluster management layer, enabling API-based cluster operations and node self-healing.
• Designed and implemented API-based network configuration such as bonding, VLANs, routing & failover.
• Utilized Salesforce to provide an integrated asset tracking system for operations, sales & support.

Senior Software Engineer – Kernel

• Designed and built kernel-based tracing tool enabling complete system-level overview during abnormal events such as high latency, lock contention and application sluggishness.
• Developed an IPVS scheduler enabling consistent hashing of network packets across load balancers.
• Prototyped networking patches to reduce network latency for latency sensitive applications.
• Built performance framework to identify inefficiencies in image fetching logic within iPhone app.

Senior Software Systems Engineer – Core OS

• Enhanced kernel and file system from 32-bit, uni-processor to 64-bit Multi-Processor aware.
• Re-architected components of the core operating system to scale across multiple processors.
• Prototyped, designed and implemented performance improvements to the file system which included CPU optimizations, disk defragmentation, file cache efficiency and disk I/O reductions.
• Refactored codebase to be highly object-oriented and unit testable. In addition, built a testing framework to enable developer tests to be run on development machines as well as target devices.
• Mentored junior team members through design reviews and technical knowledge transfer.

Senior Embedded Software Engineer

• Developed and debugged the high-end remote software platform, including areas such as NAND file system, RF4CE drivers and system middleware.
• Designed a graphics framework and scalable data model for the next generation Harmony remote platform, offering enhanced functionality at a reduced Bill Of Materials cost.
• Developed cross-platform canvas platform to develop apps across embedded devices & iPhone.

Software Engineer – iPod Touch/Nano & iPhone

iPod

• Developed 3rd party accessory communication framework for inbuilt accessories such as Nike+.
• Brought up new hardware platforms through driver and boot loader implementation.
• Optimized components of the system to cater to better power saving methodologies. Subsystems included audio engine, video engine, application frameworks, drivers and core operating system.
• Developed middleware and frameworks to support the use of new technologies for applications running on a real-time operating system.

iPhone

• Brought up iPhone embedded Bluetooth stack. Achieved Bluetooth "Best of the Breed" Award.
• Developed a cross-platform code base to enable debugging of the software stack on multiple platforms including Mac OS X, Windows, Linux and an RTOS environment.
• Worked closely on improving Bluetooth/WiFi coexistence to enable a high user experience.
• Designed other Apple wireless technologies such as Nike+ and the iPhone Bluetooth Headset.
• Patented: Group Formation Using Anonymous Broadcast Information — sharing anonymous "tokens" over existing wireless protocols.

WLAN & VoIP Software Developer

• Developed and debugged RTP and SIP stack protocols for the first WiFi BlackBerry devices.
• Ported drivers for a new wireless network card chipset to interact with existing code base.
• Re-designed code segments into a more modular architecture to enable rapid development.

Master of Applied Science (MASc)

• Research: RR-TM, a runtime for eager software transactional memory (C++, LLVM, x86 assembly) — designed a lower-overhead STM API, an LLVM pass rewriting shared-memory accesses into runtime calls, and a path-sensitive alias-analysis optimization.

Bachelor of Mathematics — Honors Computer Science (Software Engineering Option)