Jatin Sharma · Senior Software Design Engineer at Newgen Software
Enterprise software by day; by night, aidevkit, a package manager for AI agents, skills and MCP servers, plus SaaS and developer tools. Frontend, backend, infrastructure and AI, with a standing curiosity about the layers below: networks, operating systems and memory.
4 projects · 3 live in production
PDF · one page
React, Next.js, MUI, microfrontendsUI architecture, themes, RTL, Webpack and Vite buildsNode.js, Spring Boot, RESTAPIs, auth, enterprise integrationsMongoDB, PostgreSQL, Oracle, RedisModelling, caching, query behaviourDocker, Caddy, Jenkins, AWSDeployments, reverse proxies, CI/CDHTTP, TLS, TCP/IPWhat actually happens on the wireLinux, processes, C++, memoryWhere the abstractions bottom outthe path so far
I started with application development: components, endpoints, forms. It worked, and at some point “it works” stopped being a satisfying answer. Why is this request slow? What does the database do with that query? What actually happens between a browser and a server?
Each question pulled me one layer down: from JavaScript into backend services, into databases, into networking, then Linux and Docker. That curiosity is now pointing at C++, operating systems and distributed systems, and, more recently, at what it takes to run AI systems.
At work that habit shows up as debugging production problems and carrying large migrations through. Outside work it shows up as building the whole thing, then running it.
Jatin
Each one started from a specific problem. The pages behind them cover what I built, the decisions that mattered, and the architecture.
01 · Developer platform · AI tooling
A package manager and public hub for AI agents, skills and MCP servers: one command installs a building block into Claude Code, GitHub Copilot or Codex, and one command takes it back out.
02 · SaaS · full lifecycle
A task management app I designed, built, deployed and run myself, from the first schema to HTTPS in production.
03 · Developer tool · AI
Paste a package.json and get dependency guidance grounded in real registry and vulnerability data, with an LLM used only to explain and prioritise.
04 · Developer tools · Chrome extension + npm CLI
Two tools for problems I hit in real codebases: NetworkLogger, a Chrome extension that keeps network logs when tabs and popups close, and RTL-Migrate, a published codemod that moves CSS to logical properties for right-to-left layouts.
Newgen Software Technologies · Senior Software Design Engineer
Senior SDE.
Building NewgenONE at Newgen Software: one enterprise platform for content, process and customer communication. Promoted from SDE-2 to Senior SDE in July 2026.
NewgenONE is Newgen’s enterprise platform for banks, financial institutions, insurers and governments. It brings content management, business process management and customer communication management together, with business rules, document intelligence and AI agents built on top, and it is used by around 500 enterprises across 79 countries.
I build across its frontend, backend and platform layers: features for major releases, client requirements, production fixes and performance work, while keeping a large product modern as it keeps shipping.
ECM, BPM and CCM in one low-code platform, with business rules, document intelligence and AI agents on top.
The same product ships to public cloud, private data centres and hybrid setups.
Deployed on the Java application servers banks already run.
Runs on the databases regulated institutions standardise on.
Under the hood: Java and Spring Boot services, REST APIs, FreeMarker templates, React and MUI microfrontends, Docker and CI/CD pipelines.
Alongside features: microfrontend compatibility, theme architecture, RTL support, and platform-wide React, MUI and Java upgrades.
I learn by building. These are areas I’m working through, not claims of expertise.
Pointers, allocation and object lifetimes, to see what a garbage collector or a framework hides.
systemsSockets, framing and parsing, down to what a request is before a framework gets hold of it.
networkingHandshakes, certificates and why HTTPS works the way it does.
securityCaching, rate limiting, queues and logs, and how they behave under failure.
dataLong-lived connections and the state and scaling problems that come with them.
realtimeProcesses, scheduling and concurrency, from the shell side and the kernel side.
operating systemsNetworks, volumes, restarts, and the failure modes that appear once there is more than one machine.
infrastructureAgents, model APIs, AI-assisted engineering, and what running them costs.
AI