Ganesh Gunasegaran

AI Systems Architect · Staff Software Engineer

Pondicherry, India · India/Global Remote

me@itsgg.com itsgg.com linkedin.com/in/itsgg github.com/itsgg


Professional summary

I build AI and distributed systems, and the controls that keep them reliable. At Fluxon, I lead engineering for Project Alexandria, a privately backed initiative aiming to translate 1,000 public-domain classics into major languages and publish them free as PDFs, EPUBs and audiobooks, and contribute to architecture and delivery for the Google Sustainability Performance Hub. My work combines hands-on implementation with system design, delivery and operations across AI, fintech, payments and real-time platforms.


Core skills

  • AI Systems: Applied AI, LLM Systems, RAG, Vector Search (pgvector), Prompt Engineering, Predictive Systems
  • Architecture & Platform: System Design, Distributed Systems, API Architecture, Event-Driven Systems, Platform Modernization
  • Engineering Leadership: Technical Strategy, Cross-Functional Delivery, Mentorship, Code Quality, Operational Excellence
  • Backend & APIs: Ruby on Rails, Kotlin, Java/Spring Boot, Node.js/Express, Python/Flask, REST, GraphQL, gRPC
  • Frontend & Product: React, Vue.js, Angular, TypeScript, SPA/Micro-Frontends, Design Systems
  • Cloud & Data: PostgreSQL, MySQL, Redis, MongoDB, Docker, Kubernetes, Terraform, CI/CD, GitOps

Professional experience

Fluxon, Remote

Staff Software Engineer · Jun 2025 – Present

Key projects: Project Alexandria, Google Sustainability Performance Hub

Project Alexandria is a translation and publishing system for public-domain classics across 15 languages, with free PDF, EPUB and audiobook editions. I lead its engineering end to end: the web reader and its AI reading assistant, the translation pipeline, and the LLM spend controls and guardrails.

  • Built the LLM cost and safety layer: a spend kill switch, per-user spend attribution and daily caps, fallback between Anthropic and OpenAI with a circuit breaker, and prompt-injection defenses and topic limits for the reading assistant.
  • Extended the translation pipeline, serverless jobs on Modal that transcribe source books, translate them and rebuild each as HTML, PDF and EPUB: rejection of incomplete translation replies, repair of unreadable characters in source PDFs, language detection at ingest, and support for very large books.
  • Built reader and catalog features: progressive rendering for very large books, fuzzy search across every translated title, generated covers, and swipe page-turning on mobile.
  • Added the production safeguards: error monitoring and alerts, signed webhooks and access controls, rate limiting across instances, load and accessibility tests, CI quality and security gates, and runbooks.

Google Sustainability Performance Hub

  • Contribute to architecture and delivery with product, design and engineering stakeholders.
  • Build Angular/TypeScript interfaces that simplify complex internal workflows.
  • Develop Kotlin services and integrations that improve enterprise data-flow reliability.
  • Drive engineering standards and technical alignment to improve delivery quality.

Turing, Remote

Full Stack Architect · Feb 2022 – May 2025

Key projects: Lebara, Q.ai (Forbes), b-cube.ai, Buzz Solutions

  • Drove architecture across client projects in telecom, AI investing, and predictive analytics.
  • Lebara (Telecom): Produced impact assessments, defined modernization paths, evaluated payment/MNO/money-transfer integrations, authored API gateway blueprints, and reviewed scalability implications before rollout.
  • Q.ai (AI Investing): Led backend execution with ML teams, built real-time metrics services in Python/Flask/Pandas, cut analytics processing time by about 40%, and designed federated API patterns across Rails, Kotlin, Python, and Java/Spring Boot.
  • b-cube.ai (Trading Systems): Developed AI-assisted trading strategies and high-performance JavaScript tooling for execution workflows.
  • Buzz Solutions (Predictive Analytics): Rebuilt operator dashboards in Vue.js, split monolithic components into Node.js/Express services, trained fault-detection models in Python/PyTorch, and implemented GitOps delivery on GCP.

BukuKas (Lummo), Remote

Chief Architect · Dec 2020 – Feb 2022

  • Architected core fintech systems across campaigns, payments, and notifications in collaboration with product, growth, and operations teams.
  • Designed campaign and gamification flows that delivered >2x user participation.
  • Built a unified notification service (SMS, WhatsApp, push, WebSocket) with provider failover.
  • Led BukuKas Pay architecture for P2P and merchant transaction journeys.
  • Implemented fraud controls, OTP verification, and transaction observability.

BigBinary, Remote

Senior Architect · Aug 2019 – Nov 2020

  • Delivered core Neeto capabilities including Stripe payments and desktop tooling.
  • Built cross-platform applications using Electron, React, and React Native.
  • Implemented customizable PDF reporting for customer workflows.
  • Led design-system and component refactors to improve consistency and reuse.

MoneySmart, Bengaluru

Senior Software Architect · Jan 2018 – Jul 2019

  • Migrated insurance modules from monolith to Rails-based services.
  • Built micro-frontend SPA architecture for independent module delivery.
  • Developed reusable lead-provider connectors as Ruby Gems.
  • Created the “Rocketship” design system in Vue.js and drove shared platform standards.

Crossover (Engine Yard), Remote

Chief Software Architect · Jan 2017 – Dec 2017

  • Added platform support for Node.js and Elixir/Phoenix workloads.
  • Modernized the dashboard from Merb/Backbone to Rails + React.
  • Refactored monolithic services into modular microservices.
  • Architected billing, ledger, and addons services.

Akshi, Pondicherry

Founder · Jan 2013 – Nov 2016

  • Founded and architected a social learning platform for teachers and students, incubated at SageWork in 2013 and continued as a sabbatical project alongside my role at Pramati.
  • Built Rails backend services for real-time classroom and content workflows.
  • Developed custom adaptive-bitrate live video streaming.
  • Implemented virtual classrooms with live video, chat, and whiteboarding.
  • Scaled the platform to support 4,000+ concurrent students.

Pramati Technologies, Chennai

Architect · Jan 2014 – Nov 2016

  • Led mobility and machine-learning initiatives across client programs.
  • Delivered systems using Rails, Java/Spring, Angular, and Spark.
  • Drove open-source adoption and engineering standards.
  • Designed AI analytics workflows using SAP HANA and ML components.

SageWork, Pondicherry

Founder · Nov 2008 – Dec 2013

  • Founded and scaled a consulting firm delivering web and mobile products.
  • Led engineering teams across concurrent startup and enterprise projects.
  • Architected backend systems for complex business workflows.
  • Delivered products using Rails, JavaScript/jQuery, and Titanium Native.

RevenueMed, Atlanta

Development Manager · Dec 2007 – Oct 2008

  • Led development of medical coding and billing workflow applications.
  • Built WordPress/PHP LMS modules for coder training.
  • Developed an interactive molecular data-mining tool for research teams.

IBM/HCL, Chennai

Team Leader · Jun 2003 – Sep 2007

  • Led the team responsible for porting WebSphere MQ to Tandem NSK (C/POSIX), enabling reliable cross-platform enterprise messaging.
  • Designed and implemented a core SSL module in C/POSIX, strengthening security in enterprise application environments.

Selected independent work

vazhi (public)

  • Built and published a TypeScript LLM router for the pi coding agent that probes free providers, ranks available models nightly, routes each request and falls back when a provider fails or hangs.

Dream-RSI Local (public)

  • Reimplemented an unreleased agent-search method from its paper and ran a controlled three-seed test on a local model. It found no score improvement, traced 11 crashes across 27 generated policies and repaired 7 from tracebacks alone.

Nodi (public)

  • Built and shipped an Omarchy command bar for application, window, file and clipboard workflows, with questions routed to Claude, Codex, Gemini or user-owned agents over the Agent Client Protocol.

Kalvi (private, in development)

  • Designing and building a learning system for durable learning and verifiable evidence of ability, starting with an append-only learner record and randomized aided and unaided probes on real work.

Arivu (private, in development)

  • Building a personal agent with its own memory, permission policy and session record, exposed to Nodi, a terminal, Telegram and scheduled routines through the Agent Client Protocol. Autonomous changes pass sandbox, tamper and held-back acceptance gates.

Viveka (private, research and design)

  • Designing an agent runtime where models propose explicit claims, experiments and actions and deterministic code validates them. Its first test will compare accuracy and resistance to planted documents against a plain research agent at the same cost.

Education

Bachelor of Technology (Distinction with Honours)

Pondicherry University, Pondicherry