Skip to main content

Member of Technical Staff, Oracle

Hi there 👋

I’m Siddharth Mavani, a Member of Technical Staff at Oracle, where I build backend and distributed systems for Oracle Cloud Infrastructure (OCI). I like taking on tricky reliability and scale problems and turning them into secure, observable services.

Lately, that has meant productionizing zero-trust controls, building health-check systems that spot failures quickly and improve availability at scale, and developing AI-assisted automation and reusable agent skills that help engineers respond to critical incidents. I enjoy carrying an idea all the way from design to production— making it observable, automating the tedious bits, and building confidence with solid tests.

Away from the day job, I like building projects that I think are cool and fun. Recent projects include OrbitKV, a Dynamo-inspired distributed key-value store that explores quorums, vector clocks, tombstones, and read repair; and Aurelia Ledger, a Go/PostgreSQL double-entry ledger focused on immutable records, transactional correctness, and concurrency-safe financial workflows.

When I’m not coding, you’ll usually find me following Real Madrid, playing or watching cricket, or trying to stay consistent at the gym. I also love getting out of the city—whether that means planning a trip somewhere new or spending a weekend trekking and chasing a good view.

Send me a message

Experience

Member of Technical Staff

Oracle — Cloud Infrastructure Team (OCI)Full-time
July 2024 – Present

Bangalore, Karnataka, India

  • Co-authored and defended the ZPR (Zero Trust Packet Routing) Deny architectural design before principal/staff network architects, then productionized explicit-deny enforcement that strengthens customers’ security posture through finer-grained, least-privilege traffic controls; processes 100K+ evaluations/s globally with a one-minute regional peak of 158K evaluations/s.
  • Designed and implemented a distributed heartbeat-based health-check system for packet routing that improves availability by preventing prolonged traffic blackholing; sequence-validated probes detect peer failures within ~6–8s and reroute 1T+ packets/month (~390K/sec) to healthy paths.
  • Optimized the health-check system’s dataplane probe communication using a compact binary wire format and direct packet-buffer serialization, reducing packet overhead by ~23%.
  • Reduced operational remediation effort by ~78% by developing AI-assisted automation workflows and runbook tooling for Sev-1/Sev-2 incidents, later productized into a reusable operational AI agent skill adopted across the engineering organization.
  • Owned customer-facing metrics instrumentation for ZPR Deny, driving cross-timezone coordination with 2 US teams to secure approvals for UI/UX integration, metric whitelisting, publishing, and Grafana dashboarding.
  • Mentored an intern and a new-hire engineer on team infrastructure, codebase architecture, and CI/CD pipelines; guided the intern’s test automation initiative, reducing manual testing effort by ~73% (~10 hrs per release), now adopted as part of the team’s standard test suite.

Project Intern

Oracle — Cloud Infrastructure Team (OCI)Full-time
May 2023 – July 2023 · 3 months

Bangalore, Karnataka, India

  • Worked on an open-source large language model called GPT4All-J.
  • Optimized the LLM’s inference speed using Triton Inference Server with the FasterTransformer backend.
  • Containerized and automated the build process using Docker scripts.
  • Reduced inference time from 37 seconds to 1.7 seconds with no additional cost, increasing speed by 21×.

Full Stack Developer

Product LabsPart-time
January 2022 – April 2022 · 4 months

Hyderabad, Telangana, India

  • Worked in a team of 4 to develop an end-to-end OCR-based web application using the MERN stack.
  • Designed and developed the user authentication portal, where users were verified using JSON Web Tokens (JWT).
  • Built RESTful APIs using Node.js and Express.js for the backend server, with MongoDB for the database.
  • Containerized the application using Docker.

Projects

A focused selection of work across distributed systems, backend engineering, machine learning, and systems programming.

OrbitKV

A Distributed Key-Value Store

A leaderless Dynamo-style key-value store in Python with a configurable 32-bit consistent-hash ring, N/R/W quorum coordination, and Protocol Buffers/gRPC replica communication. Vector clocks preserve concurrent writes as sibling versions, while tombstones and read repair support eventual consistency across a real three-node REST/gRPC cluster validated by a 36-test unit and integration suite.

  • Python
  • Distributed systems
  • gRPC

Aurelia Ledger

Double-Entry Token Accounting Service

A Go/PostgreSQL double-entry ledger service with an immutable journal, database-enforced balancing, rebuildable projections, reconciliation, and paginated history. Concurrency-safe deposit, spend, adjustment, and reservation workflows use atomic transactions, deterministic row locking, scoped idempotency keys, and overdraft protection, backed by a 30-test domain, HTTP, and PostgreSQL integration suite.

  • Go
  • PostgreSQL
  • REST APIs

LoCoML

A Low-Code Machine Learning Platform

A low-code machine-learning platform for streamlined model training and analysis, with model versioning, data preprocessing, and automated machine learning workflows.

  • Machine learning
  • AutoML
  • Low-code

CineStream

A Video Streaming Platform

A microservice-based video streaming platform with flexible subscriptions, personalized recommendations, and a workflow for requesting new movies.

  • Microservices
  • Streaming
  • Full stack

Linux C Shell

A Unix-Style Command-Line Shell

A Linux shell implemented in C with built-in commands, job control, process information, input/output redirection, pipelines, and signal handling.

  • C
  • Linux
  • Systems programming