Garvit Galgat

WORK

Selected engineering work

ORACLE · MEMBER OF TECHNICAL STAFF

Oracle Unified Directory

About Oracle Unified Directory

SINCE JUN 2023

I work on a distributed directory platform used in enterprise deployments, spanning performance, testing, security, concurrency, and replication.

  • Query performance. Improved throughput by 25% for deployments with 10M+ entries through indexing, caching, and filter-evaluation changes.
  • Test automation. Built a functional test-generation and orchestration framework using AI-assisted rule synthesis, tool-calling workflows, and automated validation.
  • Security. Remediated multiple CVSS 8.0+ vulnerabilities in OUD’s core code. Developed controlled proof-of-concept exploits to assess exploitability and verify fixes for selected vulnerabilities.
  • Release and distributed systems. Contributed to the OUD 14c release across engineering teams and resolved concurrency, memory-management, and replication issues in distributed deployments.
ENGINEERING AREAS
Query performance · indexing and caching · concurrency · replication · security research · test automation · release engineering
TECHNOLOGY
Java · LDAP · Python · MCP-based LLM tooling

PANOPH · END-TO-END PRODUCT ENGINEERING

An ophthalmology learning platform, built from the study screen to the tools that keep it running.

Visit PanOph

LAUNCHED MAR 2026

HOW IT WORKSFrom a reviewer’s knowledge to a student’s study screen.
  1. Reviewers create content

    Reviewers prepare lessons and questions. Contributors add teaching images.

    Review app + Image Vault · Next.js

  2. Content is published

    Reviewed content is released to the shared library, ready for students to study.

    Workers API · R2 content storage

  3. Students study

    Students read, practise, and return to their learning on web or mobile.

    Web app · Next.js / PWA · Flutter mobile app

    • Sign inOne identity across apps · Firebase Auth
    • Access a subscriptionWeb payments · Razorpay
      Mobile purchases · RevenueCat
    • Keep learning progressProgress and access records · D1 database
  • Cross-platform learning. Built Next.js/PWA and Flutter apps where students study topics, take assessments, work through clinical scenarios, and save content. Both apps share accounts and learning progress. Approximately 100 daily active users as of September 2026.
  • One access model. Connected Razorpay web subscriptions and mobile-store purchases to one access system. It verifies signed payment notifications, stores provider records, and checks that payments and access agree, with tools to investigate and recover from failures.
  • Safe publishing. Built tools to review, approve, and publish versioned content. Teaching images go through consent and deidentification checks before publication.
  • Production operations. Built admin tools to investigate payment failures, resolve content reports, monitor system health, and track learning activity, referrals, and creator performance.
  • Paid product. Approximately ₹5 lakh+ in revenue from launch in March through September 2026.
ENGINEERING AREAS
Cross-platform product engineering · identity and access · payments and reconciliation · event-driven reliability · content versioning and publishing · production operations
TECHNOLOGY
Next.js · TypeScript · Flutter · Hono · Cloudflare Workers, D1, R2, KV, and Queues · Firebase · Razorpay · RevenueCat
LEARNER EXPERIENCE
Web and mobile apps sharing accounts, content, progress, and subscriptions
PLATFORM SYSTEMS
Content publishing, teaching images, payments, analytics, and support
SCALE · SEP 2026
2,000+registered accounts150+paid customers

Read the engineering story

TELECRAFT LABS · LEAD DEVELOPER

A multilingual text-to-video generation platform.

JAN 2025FEB 2025

Built in January–February 2025, before the GPT Image API arrived in April. Keeping a character recognizable across generated scenes was a central challenge. I built a pipeline that reused character context and combined generated images, multilingual narration, and video composition.

HOW IT WORKSFrom a written story to a finished video.
  1. Submit a story

    The user chooses a story, language, voice, and visual style.

    Web app · Firebase sign-in

  2. Generate scenes & narration

    The story becomes scene images and spoken narration, with character details reused between scenes.

    API Lambda · ImagineArt · ElevenLabs

  3. Assemble the video

    Images and narration are combined with music and transitions.

    VideoGen Lambda · FFmpeg

  4. Get the finished video

    The video is saved and its link is attached to the user’s project.

    S3 video storage · Project database

  • Story continuity. Structured stories into scenes with reusable character descriptions and visual context to keep characters recognizable between generated scenes.
  • Media pipeline. Combined generated visuals, ElevenLabs narration in 10+ languages, and FFmpeg composition so one story submission could produce a finished video.
  • Infrastructure. Migrated the workload from an always-on EC2 deployment to event-driven Lambda and S3.
  • Operations. Added CloudWatch monitoring and published 80+ generated videos on YouTube, accumulating 15K+ views.
ENGINEERING AREAS
Character continuity · scene orchestration · multilingual narration · media composition · serverless architecture · observability
TECHNOLOGY
Python · LangChain · FFmpeg · ElevenLabs · ImagineArt · AWS Lambda, S3, and CloudWatch
INFRASTRUCTURE COST
95% lower
MONTHLY COST
$120 → $6
PUBLISHED VIDEOS / VIEWS
80+ / 15K+

Read the engineering story