SAIKIRAN KALLURI

Product work and UI bug fixesimprovement

Started in the existing Aude product — a Next.js frontend with a Node and Express backend using Redis — exploring the codebase and fixing bugs across it, mostly on the UI. This first month built up the familiarity with the product's data flows that the diagnostics work later depended on.

Technologies: Next.js · React · Node.js · Express · Redis

Internal diagnostics dashboardfeature

After the first month I was given the chance to design and build an internal diagnostics application, which I built as a Next.js monolith. It talks directly to the database, checks whether the main app is generating correct data for every client, and shows the state of all of them in one dashboard. The interface is built for the engineers who debug it: what is not working is visible at a glance, an investigation can be deep-linked and shared as a URL, and a user's-eye view shows the UI as the user sees it next to the API response behind it.

Project: Aude Diagnostics – Internal Data-Integrity Dashboard

Technologies: Next.js · React

Deployment behind the existing single sign-oninfrastructure

Containerised the diagnostics app with Docker and deployed it on AWS, serving it from the same origin as the main application so it reuses the existing single sign-on instead of introducing a second authentication system. Access is further restricted to authorized members of the engineering team.

Project: Aude Diagnostics – Internal Data-Integrity Dashboard

Technologies: Docker · AWS · Next.js

Skeleton loading statesimprovement

Took part in UX improvement discussions and proposed skeleton loading for the app, which had no loading view. The proposal was approved, I implemented it, and it is now in production.

Technologies: Next.js · React

Digichikitsak — voice-enabled prescriptions and AWS deploymentfeature

A MERN web application implementing electronic prescriptions: creating new ones, converting existing records, and storing them securely. I implemented voice input and commands using the ElevenLabs and Bhashini APIs, then containerised it with Docker and deployed the project to AWS ECS — setting up the infrastructure through to CloudWatch alerts — with CI/CD through GitHub Actions.

Project: Digichikitsak – Digital Prescription Platform

Technologies: React · Node.js · Express · MongoDB · Docker · AWS · Git

Elphie — cloud video and live streaming for dental imaginginfrastructure

Elphie pairs a C++ desktop application, which drives an intraoral camera and runs a caries-detection model, with a web application. I worked inside the C++ application itself to upload the captured video to AWS S3 and publish a live stream through AWS IVS, so a patient's family or dental students can watch an examination as it happens. I also built the web application from scratch as a Next.js monolith to show recordings and carry the live stream, with AWS Cognito authentication and appointment booking — a doctor selects a patient, streams live from the C++ app, and the recording is stored against that patient automatically before the diagnosis is written up. I deployed the web application on AWS Amplify.

Project: Elphie – Dental Imaging Capture and Live Streaming

Technologies: C++ · AWS · Next.js · React

Axion — mobile-driven desktop automationfeature

A Docker-based desktop automation tool in three parts: a MERN cloud relay, a mobile-first Next.js web application, and a Python desktop application. A user types a prompt or speaks into the web app, which creates a task and pushes it to the relay; the desktop app picks it up and executes it. Every step of the run is visible from the phone, and the user can steer the task midway.

Project: Axion – Mobile-Driven Desktop Automation

Technologies: Next.js · React · Node.js · Express · MongoDB · Python · Docker

HealthyMitra — remote caregiving platformfeature

A Next.js monolith letting caregivers look after patients digitally through care circles, which users join as patient or caregiver. Caregivers book appointments and diagnostic reports for their patient and get instant updates from doctors and diagnostic centres; patients can reach doctors any time through chat with WebRTC audio and video calls. Users post updates to their circle with per-post visibility — everyone, caregivers only, doctors, or a selected few. I also implemented Razorpay payments and an admin dashboard covering the platform, and containerised it with Docker for deployment on AWS ECS with CI/CD through GitHub Actions.

Project: HealthyMitra – Care Circles for Remote Caregiving

Technologies: Next.js · React · WebRTC · Razorpay · Docker · AWS · Git

AWS Cloud curriculumresearch

Completed a three-month virtual internship focused on AWS Cloud technologies, delivered through a collaboration between Eduskills and AWS Academy.

Technologies: AWS

Meeting session

Participants join a WebRTC-based meeting and interact through the browser.

Private inference

TensorFlow.js performs engagement inference on the client so raw video does not need to be uploaded for analysis.

Instructor insights

Derived session signals are summarized for instructors and can be explored through a retrieval-augmented natural-language workflow.

Model development

Developed and evaluated a hybrid ensemble using a 1D CNN and 1D ResNet with the DAiSEE dataset.

Browser integration

Moved inference into TensorFlow.js and integrated it with the meeting experience.

Analytics workflow

Connected engagement results to instructor visualizations and a retrieval-augmented query flow.

Balancing useful engagement analytics with participant privacy.

Kept video processing in the browser and shared derived signals instead of raw frames.

Lesson: System architecture can make privacy a default property rather than an optional policy.

Record management

Users create and manage structured certificate records through the web interface.

Data storage

The backend persists application data in DynamoDB.

Media handling

Certificate images are uploaded and delivered through Cloudinary.

Team planning

Split the platform into clear implementation areas and coordinated work across a four-member team.

Full-stack implementation

Connected the React interface to Node.js and Express services backed by DynamoDB and Cloudinary.

Deployment

Prepared and deployed the application for institutional use.

Coordinating application data with separately hosted certificate media.

Stored structured records in DynamoDB while using Cloudinary for media-specific storage and delivery.

Lesson: Different data types benefit from services designed for their access and delivery patterns.

Match context

The application collects runs, wickets, overs, recent performance, and team information.

Model inference

A Keras model processes the current context and predicts the final score.

Feature preparation

Selected and prepared match-state inputs that influence an innings' final score.

Model development

Built and evaluated a deep neural network with Keras.

Prediction interface

Connected model inputs and output to a practical score-estimation experience.

Representing a changing match in a form a model can use consistently.

Combined current score information with recent-performance and team context.

Lesson: Feature design is as important as model architecture for applied prediction problems.

Guided input

Users provide the experience and background needed to create résumé content.

AI generation

GPT-4 Turbo transforms the input into ATS-oriented résumé material.

Paid access

Razorpay handles the application's payment flow.

Product flow

Designed a guided sequence from user input to generated résumé output.

AI integration

Connected structured prompts and application data to GPT-4 Turbo.

Payment integration

Added Razorpay to support paid access.

Goal planning

Users define goals that provide context for daily work.

Priority decisions

Tasks are organized with the Eisenhower Matrix according to urgency and importance.

Daily review

The application helps users review daily productivity and progress.

Workflow design

Connected goals, tasks, prioritization, and review into a coherent mobile flow.

Flutter implementation

Built the application interface and behavior with Flutter and Dart.

Keeping long-term planning useful during daily task decisions.

Connected goals directly to daily planning and made prioritization part of the core workflow.

Lesson: Productivity tools work best when planning levels reinforce one another.

Content layer

Every project, role, and skill is a typed JSON file, read and validated at build time through server-only modules — the same data renders the case-study pages, the skill graph, and the downloadable résumé and CV.

Motion layer

GSAP and ScrollTrigger drive a pinned projects carousel with one single source of truth for scroll position; a Lenis-smoothed scroll and a canvas starfield sit underneath the rest of the site.

Verification layer

GitHub Actions runs lint, type checks, unit tests, and a Playwright suite against a production build on every change, before anything reaches this page.

Measuring before fixing

The projects grid was dropping to 83ms a frame on an iPad. Rather than guess at the cause, isolated each suspect in turn until backdrop-filter across every card turned out to be the entire cost, and fixed only that.

Testing what a unit test cannot see

Added a Playwright suite that runs against a real production build, because the regressions this site actually had — a sideways-scrolling page, a document that prints blank mid-animation — are invisible to jsdom. It found a real bug on its first run: a full-screen dialog's Close button sitting under the fixed site header, so a click on Close was landing on the audio toggle instead.

Writing down what is unfinished

Keep a status document recording known defects, deferred work, and the reasoning behind decisions like pinning a dependency — so the gaps are visible rather than something a visitor has to find themselves.

A backdrop-blur effect across the projects grid dropped frame times to 83ms on an iPad, and the obvious guesses about the cause were wrong.

Measured instead of guessing, and isolated the cost by removing one suspect at a time until backdrop-filter across every card turned out to be the entire cost.

Lesson: Measuring the actual page beats reasoning about what should be slow.

A dependency update silently removed six icons the site's skill map depended on, and the build only caught one of them before stopping.

Checked every imported icon against the new version rather than trusting the first error, found five more missing, and pinned the dependency with the reasoning written next to the code that depends on it.

Lesson: A dependency bump is a claim, not a fact, until it has actually been checked.

React

A component-based library for building interactive user interfaces.

How I use it: I use React for interactive application interfaces, shared UI components, and client-side state.

Next.js

A React framework for server-rendered, statically generated, and full-stack web applications.

How I use it: I use it for route-based applications, server and client component composition, metadata, and APIs.

JavaScript

The programming language that powers interactive behavior across the modern web.

How I use it: I use JavaScript for application logic, asynchronous workflows, APIs, and browser interactions.

TypeScript

JavaScript with static types for safer, more maintainable applications.

How I use it: I use it to model application data, component contracts, APIs, and reusable utilities.

Node.js

A JavaScript runtime for servers, tooling, and backend applications.

How I use it: I use it for server logic, API integrations, data access, and development tooling.

Express

A minimal Node.js framework for HTTP APIs and backend services.

How I use it: I use it to define routes, middleware, validation, and service integrations.

Tailwind CSS

A utility-first CSS framework for composing responsive interfaces.

How I use it: I use it for layout, responsive behavior, theming, interaction states, and rapid UI iteration.

HTML5

The semantic structure and content layer of web interfaces.

How I use it: I use semantic elements, accessible labels, forms, and document structure in every web project.

CSS3

The layout, visual styling, and responsive presentation layer of the web.

How I use it: I use modern layout systems, responsive queries, custom properties, and transitions.

WebRTC

Browser technologies for real-time peer-to-peer audio, video, and data.

How I use it: I use it to establish real-time participant media sessions in the browser.

Python

A general-purpose language used for machine learning, data work, and backend automation.

How I use it: I use it for model development, data preparation, experimentation, and automation.

Java

A strongly typed object-oriented language used across application and backend development.

How I use it: I use Java to practice structured programming, object modeling, and algorithmic problem solving.

TensorFlow.js

A JavaScript machine-learning library for training and inference in browsers and Node.js.

How I use it: I use it for privacy-first client-side model inference.

LangChain

Tools for composing language-model applications, retrieval, and data-aware workflows.

How I use it: I use it to organize retrieval and language-model interactions around application data.

Retrieval-Augmented Generation

A pattern that grounds language-model responses in retrieved application data.

How I use it: I use retrieval to connect user questions with relevant structured or indexed project data.

Deep Learning

Machine-learning techniques based on multi-layer neural networks.

How I use it: I use neural networks for engagement classification and score-prediction problems.

Keras

A high-level neural-network API for building and training machine-learning models.

How I use it: I use it to define model architectures and run training experiments.

C++

A systems language used here for a desktop application driving camera hardware and cloud video pipelines.

How I use it: I used it to add cloud capability to that application: uploading captured video to AWS S3 and publishing a live stream through AWS IVS from inside the C++ application itself.

MongoDB

A document database for flexible JSON-like application data.

How I use it: I use it for application records whose structure fits a document model.

MySQL

A relational database for structured application data and SQL queries.

How I use it: I use it when application data benefits from relations, constraints, and transactional consistency.

PostgreSQL

A powerful relational database with rich SQL and data-integrity features.

How I use it: I use it for structured data, relationships, constraints, and reliable querying.

Supabase

A backend platform built around PostgreSQL, authentication, storage, and APIs.

How I use it: I use it to prototype and build database-backed application features quickly.

DynamoDB

AWS's managed NoSQL database designed for predictable, scalable access patterns.

How I use it: I use it for serverless-style application data with explicit access patterns.

AWS

A cloud platform offering managed infrastructure, storage, databases, and application services.

How I use it: I use managed AWS services when a project benefits from scalable cloud infrastructure.

Docker

A container platform for packaging applications with consistent runtime dependencies.

How I use it: I use it to make local, test, and deployment environments more consistent.

Redis

An in-memory data store used for caching and fast key-value access.

How I use it: I use it where a request needs data faster than a round trip to the primary database is worth.

Git

A distributed version-control system for tracking and collaborating on source code.

How I use it: I use focused commits, branches, diffs, and history to manage changes safely.

Linux

An open-source operating-system ecosystem widely used for development and servers.

How I use it: I use shell tools, permissions, processes, and package workflows during development.

Figma

A collaborative interface-design and prototyping platform.

How I use it: I use it to explore layouts, component structure, visual hierarchy, and interaction ideas.

Canva

A visual-design platform for graphics, presentations, and communication assets.

How I use it: I use it for quick visual assets, presentation material, and content composition.

Tableau

A data-visualization platform for building interactive dashboards and reports.

How I use it: I use it to explore trends and communicate data through visual summaries.

Cloudinary

A managed platform for uploading, transforming, and delivering image and video assets.

How I use it: I use it to keep media workflows separate from structured application data.

Flutter

A cross-platform UI toolkit for building mobile applications from one codebase.

How I use it: I use it to compose mobile interfaces and application workflows from reusable widgets.

Dart

The typed programming language used to build Flutter applications.

How I use it: I use Dart for mobile application logic, data models, and widget behavior.

Mobile Development

Designing and building software experiences for mobile devices.

How I use it: I apply mobile-first workflows, reusable screens, and touch-oriented interaction design.

GPT-4

A large language model used for generating and transforming natural-language content.

How I use it: I use structured prompts and user context to generate ATS-oriented résumé content.

Razorpay

A payment platform for accepting and managing online payments.

How I use it: I use it to connect payment completion with application access.