Software Engineer III · Google

Udayan Joshi

Building large-scale Machine Learning models & optimization systems at Google — previously distributed cloud data security at Theom.ai.

At Google, I develop and deploy production Machine Learning models and experimentation pipelines to optimize Shopping Ad offer and format selection at scale. Previously, as a Member of Technical Staff at Theom.ai, I built distributed Go microservices and ML threat-detection systems (co-inventing 8 granted U.S. patents). I hold an M.S. in Computer Science from UC San Diego and a B.Tech. in CSE (Honors) with a Minor in Mathematics from IIT Bombay.

Get in Touch View Patents (10)
Portrait of Udayan Joshi Mountain View, CA
8 Granted
U.S. Patents (+ 2 Published Apps)
5+ Years
Applied ML & Distributed Systems
3.98 / 4.0
M.S. CS GPA · UC San Diego
AIR 90
B.Tech CS (Hons.) · IIT Bombay
Career

Work Experience

Building large-scale machine learning ranking & optimization systems, rigorous online experimentation, and resilient cloud infrastructure.

Member of Technical Staff

Theom.ai
May 2021 – Mar 2025 San Jose, CA
  • Distributed Microservices Architecture: Designed, developed, and scaled core backend microservices in Go powering an enterprise cloud data security and governance SaaS platform.
  • End-to-End Posture & Threat Pipeline: Built key subsystems across the product lifecycle—scanning customer cloud data environments, exporting structured metadata, computing risk & CIA-triad impact insights, and surfacing real-time alerts and dashboards.
  • Low-Latency Inter-Service RPC: Engineered lightweight service-to-service communication using gRPC and Protocol Buffers (protobuf) to maximize throughput and reduce tail latency across internal services.
  • Cloud-Native Infrastructure & Persistence: Deployed highly available workloads on AWS (EKS, Lambda, S3) with Docker & Kubernetes, and architected multi-tier persistence layers across Snowflake, PostgreSQL, and Redis for sensitive telemetry and metadata.
  • Applied ML for Data Security: Integrated Named Entity Recognition (NER) for sensitive data classification and anomaly detection engines to proactively uncover anomalous queries, shadow clones, insider risks, and data exfiltration.
  • Engineering Mentorship & IP: Co-inventor on 8 granted U.S. patents and 2 published patent applications in cloud data warehouse & data lake security; mentored engineering interns on Go microservices design and production software engineering.
Go gRPC & Protobuf Kubernetes (AWS EKS) AWS Lambda & S3 Snowflake PostgreSQL Redis Docker NER & Anomaly Detection

Machine Learning & NLP Intern

San Diego Supercomputer Center (SDSC), UC San Diego
Jun 2020 – Sep 2020 San Diego, CA
  • Scientific Document & Sequence Parsing: Built automated extraction pipelines to parse genetic sequences and domain metadata across heterogeneous XML, PDF, and Word scientific corpora under the supervision of Dr. Mai Nguyen.
  • Relation Extraction & Knowledge Curation: Designed an end-to-end Relation Extraction (RE) pipeline coupling Named Entity Recognition outputs with Transformer-based RE models and interactive graph visualizations—contributing to the peer-reviewed Synthetic Biology Knowledge System published in ACS Synthetic Biology.
  • Graph Representation Learning: Researched applications of Graph Neural Networks (GNNs) for automated scientific knowledge graph construction and link prediction.
Python PyTorch Transformers Graph Neural Networks SpaCy Relation Extraction

Machine Learning Intern

Veritas Technologies LLC
Summer 2018 Pune, India
  • Developed topic modeling, sequence mining, and deep learning prototypes for enterprise telemetry and log analysis under the guidance of Niranjan Pendharkar.
Python Deep Learning Topic Modeling Sequence Mining
Intellectual Property

Patents & Filings

Co-inventor on 8 granted U.S. patents and 2 published U.S. patent applications assigned to Theom, Inc., covering cloud data warehouse security, data lake threat detection, and automated risk grading.

US 12,737,335 B2 Granted · Sep 2026

Methods and Systems for Securing Data Cloning and Sharing Options on Data Warehouses

Determines primary versus cloned data warehouse assets using log analysis, lineage timestamps, and metadata fingerprinting to enforce security controls across zero-copy clones and data shares.

US 12,670,273 B2 Granted · Jun 2026

Methods and Systems for Detecting Hiding and Data Intelligence Gathering in Data Lakes and Cloud Warehousing

Detects stealthy adversary staging, column anomalies, and unauthorized intelligence collection inside cloud data lakes and warehouses, paired with dynamic masking defenses.

US 12,634,323 B2 Granted · May 2026

Methods and Systems for Attack Generation on Data Lakes

Provides an automated adversarial mimicry framework simulating reconnaissance and infiltration attack chains against enterprise data lakes to validate detection coverage and posture resilience.

US 12,561,429 B2 Granted · Feb 2026

Methods and Systems for Automatic Grading, Impact Analysis and Mapping to the CIA Triad

Quantifies datastore business value and maps live attack progression to Confidentiality, Integrity, and Availability (CIA) impact scores for prioritized security remediation.

US 12,513,163 B2 Granted · Dec 2025

Methods and Systems for Detecting Data Abuse and Data Exfiltration in Data Lakes and Cloud Warehouses

Identifies covert command-and-control (CnC) channels, anomalous bulk staging, and malicious compressed artifacts to prevent data abuse and exfiltration across enterprise data clouds.

US 12,355,796 B2 Granted · Jul 2025

Locating Shadow Vulnerable Datastores for Cloud Data Table/API/Data Lake Stores

Discovers unmanaged or cloned shadow datastores across cloud database, API, and data lake environments, computes posture drift against primary stores, and generates differential remediations.

US 12,294,589 B2 Granted · May 2025

Cloud-Platform Push for Known Data Breaches

Maps Tactics, Techniques, and Procedures (TTPs) and kill chains of emerging data breaches into proactive cloud-platform security checks pushed directly to customer data environments.

US 12,289,331 B2 Granted · Apr 2025

Methods and Systems for Locating Anomalous Query Activity on Data Stores

Uses an atypical query analysis engine over cloud databases and SaaS data stores to model access patterns, user roles, and query semantics—flagging anomalous data access in real time.

US 2024/0106847 A1 Published · Mar 2024

Methods and Systems for Detecting Reconnaissance and Infiltration in Data Lakes and Cloud Warehouses

Monitors SaaS and cloud-native data stores from inside the data plane to detect early-stage reconnaissance queries, privilege probing, and infiltration progression.

US 2023/0334162 A1 Published · Oct 2023

Minimizing Data Governance to Improve Data Security

Measures access over-provisioning across cloud data warehouses and synthesizes right-sized, context-aware least-privilege policies without heavy manual governance overhead.

Scholarship & Applied ML

Publications & Research Projects

Selected peer-reviewed research and applied machine learning collaborations across scientific NLP, climate/wildfire resilience, and sparse signal recovery.

Journal Publication · 2021

Synthetic Biology Knowledge System

ACS Synthetic Biology · Vol. 10, Issue 9

Co-authored peer-reviewed platform paper integrating text-mining, Named Entity Recognition, and Relation Extraction pipelines over scientific literature to accelerate discovery in synthetic biology.

Applied ML · UC San Diego × SDG&E

WIFIRE Vegetation & Power-Line Outage Prediction

San Diego Supercomputer Center / WIFIRE

Collaborated with San Diego Gas & Electric (SDG&E) to build predictive machine learning models forecasting power-line failures and outages triggered by vegetation and natural phenomena to help prevent California wildfires.

B.Tech. Honors Thesis · IIT Bombay

Greedy Algorithms for Non-Negative Non-Linear Sparse Regression

Advised by Prof. Ajit Rajwade

Designed, implemented, and theoretically analyzed greedy pursuit algorithms for non-negative, non-linear sparse signal recovery with applications in compressive sensing and computer vision.

Academic Foundation & Tooling

Education, Teaching & Technical Stack

University of California, San Diego

Sep 2019 – Mar 2021
Master of Science in Computer Science · GPA: 3.98 / 4.00

Focus on Machine Learning, Computer Vision, and Statistical AI. Served as Graduate Teaching Assistant for Computer Vision, CSE 250A (Graduate Probabilistic Reasoning & Learning with Prof. Lawrence Saul), and CSE 150A (Intro to AI: Probabilistic Models).

Indian Institute of Technology Bombay

Jul 2015 – May 2019
B.Tech. in Computer Science & Engineering (Honors) · Minor in Mathematics

All India Rank 90 in JEE Advanced. Served as Undergraduate Teaching Assistant for CS 101 (Computer Programming & Utilization with Prof. Ganesh Ramakrishnan) and MA 105 (Calculus with Prof. A. K. Pani).

Languages

Go (Golang) Python SQL C / C++

Distributed Systems & Cloud

Microservices gRPC Protocol Buffers Kubernetes Docker AWS EKS AWS Lambda AWS S3

Data & Security Infrastructure

Snowflake PostgreSQL Redis Cloud Data Lakes DSPM & Threat Detection

Machine Learning & Experimentation

Ads Ranking & Optimization Online A/B Testing PyTorch Named Entity Recognition Anomaly Detection Graph Neural Networks
Personal

Beyond Engineering

Outside of distributed systems and data security, I play the bansuri (Indian bamboo flute) and love listening to Indian Classical Music—always happy to trade recommendations for memorable concerts and compositions. You'll also find me on the football (soccer) pitch or occasionally revisiting algorithmic puzzles on Codeforces, CodeChef, and Project Euler.

“Err and err and err again, but less, and less, and less.” — Piet Hein, Grooks