Harsh Sharma

Frontend Engineer · Mohali, Punjab, India

I'm a software engineer with 3+ years building production web applications in TypeScript, React and Next.js. Most of my work lives in the layer between a design system and a product team: reusable component architecture, schema-driven form systems, analytics dashboards, and multilingual content tooling. I care about the abstractions that stop the same UI being written five times. I also work in Go, PostgreSQL and Docker, mostly through production integrations and side projects.

Projects

Autognosis MCP Server

A remote MCP server that exposes my professional profile as callable tools.

An AI assistant asked about my background has to scrape a page and infer the rest. There is no structured, authenticated way for it to simply ask.

Six MCP tools and an about:// resource served over Streamable HTTP, running as a Cloudflare Durable Object so every client gets a stateful session without hand-wiring transport. Access runs through OAuth 2.1 with dynamic client registration and PKCE, so a client self-registers and authenticates with no manual setup on either side.

TypeScriptMCP SDKCloudflare WorkersDurable ObjectsWorkers KVOAuth 2.1Zod

Form-Flow

A schema-driven form builder with type-safe field configuration.

Business forms get rewritten from scratch every time the requirements shift, and the validation, rendering and submission logic end up tangled together.

Separated validation, rendering, submission and UI primitives into a modular foundation, and modelled onSubmit/onSettled with TypeScript generics so wiring a custom workflow is checked at compile time rather than discovered at runtime.

TypeScriptReactReact Hook FormZod

Candidate Ranking Engine

An explainable Go ranker that scores a 100,000-candidate pool without a model.

Ranking candidates with an embedding model is quick to build and hard to defend — nobody can say why a profile landed where it did, and keyword stuffing quietly wins over demonstrated work.

Eight weighted components — career fit, retrieval and ML evidence, seniority, company fit, engagement, location, education and penalties — scored over normalized text groups, with career-description evidence weighted above the raw skills list. Penalties cover the down-rank cases, including profiles claiming more years in a skill than they have in total. Every rank carries generated reasoning grounded only in facts present in the profile. Built as a submission for an AI-engineering hiring challenge.

  • Ranks a 100,000-candidate pool in ~22 seconds, CPU-only, with no network calls
  • Go standard library only — no third-party dependencies
GoDockerGitHub Actions

Tailwind v4 Component Package

A monorepo spike on shipping a Tailwind v4 component library to outside consumers.

Tailwind v4 moved theming into CSS, which breaks the usual way a component package ships styles: import Tailwind inside the package and the consumer gets it twice, leave it out and the components arrive unstyled.

The package keeps Tailwind out of its own stylesheet and exposes only source directives and theme tokens, so the consuming app runs Tailwind once across both its code and the package sources. The stylesheet is copied into the build output and imported alongside Tailwind, which makes the same package work inside the monorepo and in an external project.

TypeScriptTailwind CSS v4TurborepoViteNext.js

Datastar Signal Generator

A Go utility that removes the manual wiring from Datastar signals.

Server-driven UI flows in Datastar need signals named and managed by hand, which drifts inconsistent across templated interfaces.

Built a lightweight Go generator that produces consistent naming and management abstractions for signals, so templated reactive interfaces stay maintainable as they grow.

GoDatastar

Go + Templ Server Sandbox

A learning sandbox for building server-first web apps in Go, without a SPA.

I wanted to work through sessions, auth, migrations and reactive UI in a server-rendered stack rather than reaching for React by reflex.

Users, sessions, password reset and gallery CRUD built from Templ components and Datastar signals, with Goose migrations embedded through Go's embed package and run on startup, CSRF protection on form actions, and Docker Compose plus Air for the local loop. It is a record of how the pieces fit together rather than a production starter.

GoTemplDatastarPostgreSQLGooseDocker ComposeDaisyUI

Experience

WebsmithsSDE 2

October 2025Present · Mohali, Punjab, India

  • Architected the frontend foundation for Auto-Connex, a multi-product automotive SaaS platform, using Next.js, TypeScript, React Query and Zustand across 4+ internal products.
  • Built reusable UI abstractions for tables, layouts, dashboards and data-heavy workflows, cutting repeated UI implementation by ~40%.
  • Launched the schema-driven form builder powering the Infinity product, letting businesses configure 100+ workflows and application forms without developer involvement.
  • Created reusable field, validation and layout patterns with React Hook Form and Zod, reducing manual form setup effort by ~60%.
  • Engineered a visual builder for trigger-based email automation, cutting campaign setup time by over 75%.
  • Integrated iframe-based Voice Agent data collection into external automotive websites over webhooks, spanning member management, subscriptions and cross-product analytics.

Logicease Tecno SolutionsSoftware Engineer

April 2023October 2025 · Ambala City, Haryana, India

  • Delivered React content creation tools for Binogi's education platform, letting educators manage structured lesson material across subjects, regions and languages.
  • Automated AI-assisted translation and localization workflows, removing repetitive manual translation work from country-specific lesson delivery.
  • Engineered a React + Remotion video pipeline with a drag-and-drop timeline editor for multilingual educational assets, with configurable banners, timing controls and per-language variations.
  • Built licensing and access management for consumers, schools and institutions — license creation, allocation, usage tracking and organization-level distribution.
  • Standardized analytics behind Segment and Mixpanel abstraction layers, improving event consistency, attribution and deep-linking visibility platform-wide.

Skills

Frontend
TypeScriptJavaScriptReactNext.jsReact QueryZustandRemotionTailwind CSSChakra UIStorybook
Architecture
Frontend architectureReusable componentsComponent librariesDesign systemsState managementAPI integration
Forms & validation
React Hook FormZodSchema-driven formsConfigurable form builders
Backend
GoNode.jsPostgreSQLREST API designWebhooks
Cloud & DevOps
DockerCI/CD pipelinesAWS S3Cloudflare WorkersFirebase
Testing
VitestPlaywrightCypressUnit testingIntegration testing
AI & SDKs
Vercel AI SDKAnthropic SDKOpenAI SDKMCP
Tooling
GitpnpmTurborepoTemplDatastarHTMX

Education

B.E. Computer Science and Engineering

Ambala College of Engineering & Applied Research (KUK) · 2023 · GPA 7.79

Diploma in Mechanical Engineering

Govt. Polytechnic Ambala City (HSBTE) · 2020 · 75.25%

Contact

I'm open to conversations about frontend architecture, design systems and product engineering roles.