Backend-focused engineer with 5+ years of experience building real-time, event-driven systems and scalable SaaS platforms. Proven track record of designing microservices, optimizing system performance (up to 65% reduction in DB load), and delivering end-to-end products from architecture to deployment. Experienced in Java, Spring Boot, and distributed data systems (PostgreSQL, InfluxDB). Strong ownership mindset with experience leading 0→1 product builds.
Work experience
- Developed a dashboard with user analytics giving insights into product usage and retention.
- Built custom rules functionality on forms from the ground up, enabling flexible automation for users.
- Developed and tested anti-malware scanning for file uploads.
- Automated outreach emails for onboarding and re-engagement based on user interaction timing.
- Owned notification and email services using scheduling, events, and multi-criteria triggers.
- Delivered front-end, back-end, and deployment work after collaborating with product and design teams.
- Contributed to monolith-to-microservices transformation, improving scalability and maintainability.
- Built a new microservice from scratch using Spring Boot, ReactJS, and MongoDB.
- Practiced TDD in Java 17 and ReactJS for robust, production-ready code.
- Participated in architectural decisions and created class and sequence diagrams.
- Contributed to a new module in an existing web application using Java 8.
- Worked extensively with MSSQL and handled complex production queries.
- Released critical production fixes within a tight 2-hour weekly window.
- Achieved a 65% reduction in database calls through strategic legacy refactoring.
- Maintained customer communication while resolving defects and improving quality.
Freelance projects
Next.js 16
TypeScript
MDX
Tailwind 4
JSON-LD
Playwright
- Built the website for an Indian intellectual property law practice, prioritising discoverability by search engines and AI assistants, as the owner asked.
- Drove internship applications and client enquiries through organic search alone, with no paid promotion.
- Statically generated every page with full article text in the HTML, and added a CI check that fails if any route turns dynamic.
- Added JSON-LD for the practice, the founder, articles, breadcrumbs, and FAQs, plus an llms.txt index, an RSS feed, and per-page share images.
- Set up Google Search Console and Bing Webmaster Tools, submitted sitemaps, and requested indexing for key pages at launch.
- Made content owner-editable as Markdown files with a template and a plain-language guide, validated at build time so a mistake fails the deploy.
- Kept all copy informational to comply with Bar Council of India advertising rules, with a disclaimer shown on first visit.
- Covered loaders, components, structured data, and generated routes with 17 Vitest files and 7 Playwright specs.
Spring Boot
InfluxDB
PostgreSQL
AWS
React
RTSP
Raspberry Pi
- Designed and built a production-ready real-time intrusion detection system integrating fibre optic sensing hardware with cloud and edge infrastructure.
- Developed an event-driven backend using Spring Boot, PostgreSQL, and InfluxDB to ingest and process high-frequency sensor data.
- Implemented edge + cloud architecture with on-site Raspberry Pi nodes and AWS-based notification services.
- Built a multi-channel alerting system (WhatsApp, Telegram, Email, ntfy) for real-time incident notifications.
- Engineered live CCTV streaming pipeline (RTSP) and integrated video feeds into a React-based monitoring dashboard.
- Created an interactive map-based UI to visualize intrusion events and system status in real time.
- Designed user and notification management systems with secure authentication.
- Collaborated directly with hardware engineers to align software with fibre optic sensing modules.
- Led a team of developers and drove end-to-end delivery from requirements to on-site deployment.
Python
PostgreSQL
React
RBAC
OpenAI
Multi-tenant
- Led architecture and development of a multi-tenant, modular SaaS platform enabling businesses to digitize end-to-end operations.
- Designed a plug-and-play module system allowing customers to enable purchased features and build custom workflows.
- Built core systems for sales, purchasing, inventory, and data management to replace manual processes.
- Implemented fine-grained authentication and role-based access control for multi-level organizational usage.
- Developed dynamic dashboards driven by module-level data and business workflows.
- Engineered modules for real-time pricing control and automatic propagation to sales systems.
- Created validation and approval layers for transactional oversight on orders.
- Delivered multi-language support and intuitive UX for operational users.
- Integrated analytics features including forecasting, segmentation, and ROI insights.
- Built AI-powered automated email classification and response drafting to prevent missed enquiries.
- Led cross-functional delivery across backend (Python), frontend (React), and database design (PostgreSQL).
- Collaborated closely with business stakeholders to translate requirements into scalable software systems.
Python
OpenAI API
REST APIs
Google Drive API
Swagger
- Designed and developed a Python-based API platform for document management and AI-driven content generation.
- Integrated OpenAI APIs for document summarization and dynamic text generation workflows.
- Built RESTful APIs with Swagger-based interface for seamless developer testing.
- Implemented session-aware paginated response handling for large AI-generated outputs with aggregation and persistence.
- Developed document lifecycle features including creation, retrieval, filtering, and storage.
- Enabled contextual summarization based on user-supplied inputs like name and location.
- Integrated external storage workflows with Google Drive APIs.
- Delivered a backend-driven system focused on API usability and extensibility.
InfluxDB
Grafana
Python
Telegram
ntfy
- Developed an early real-time monitoring and intrusion detection prototype using InfluxDB and Grafana.
- Built a live analytics dashboard visualizing sensor data, intrusion events, and system metrics.
- Implemented geospatial visualization for real-time intrusion tracking.
- Designed a multi-layer alerting system combining Grafana native alerts with custom Python notifications.
- Enabled live data ingestion and querying for time-series analysis.
- Delivered a lightweight, rapidly deployable system to validate feasibility and client needs.
Personal projects
Next.js 16
React 19
TypeScript
TensorFlow.js
Cloudflare R2
Playwright
- Built a narrative wedding site and photo platform from a 196GB photographer delivery, designed end-to-end to run on a free object-storage tier.
- Wrote a seven-stage offline pipeline that hashed 7,504 JPEGs, removed 2,468 near-duplicates, and compressed the 5,036 survivors to 1.31GB of AVIF.
- Ran face detection and recognition locally, finding 17,731 faces and clustering them into 92 people so guests can browse every photo they appear in.
- Kept all machine learning offline with locally bundled model weights — no API keys, no network calls, and no per-image cost at any volume.
- Server-rendered every gallery page from manifests so images begin downloading from the HTML rather than after a client fetch.
- Built a local-only curation tool, 404ing in production, for correcting clustering mistakes without shipping an admin surface or building auth.
- Held the app runtime to two dependencies, keeping sharp, TensorFlow, and the storage client as build-time-only tooling.
- Wrote 58 unit and integration test files plus browser end-to-end specs, every one network-safe so the suite can never generate a bill.
- Rewrote git history to remove bulk media, shrinking the repository from 2.1GB to 2.8MB while preserving 94 of 98 commits.
- Shipped a public demo from the same codebase — 1,004 photographs with every face frosted — checked by three independent face and text detectors and a build that fails if anything identifying the real wedding survives.
Next.js 16
TypeScript
Supabase
PostgreSQL
WhatsApp
Vitest
- Built a private property-management app replacing spreadsheets for tenant records, rent calculation, payment status, and payment history.
- Modelled scheduled rent increases so raises apply on their own dates instead of being remembered and re-entered each month.
- Designed a two-tier shared-PIN access model enforced server-side, chosen deliberately over real auth for a handful of non-technical family users.
- Implemented HMAC-signed thirty-day session tokens on Node's own crypto module, so rotating the admin PIN revokes every live session at once.
- Made locked access degrade rather than fail — sensitive fields are omitted and totals stay accurate, instead of the page refusing to load.
- Enforced an architecture rule keeping all business logic server-side, backed by a refactor relocating roughly ten pieces of logic out of components.
- Isolated WhatsApp broadcasts in a separate worker process so a browser-automation failure cannot take down the web app, and added a phone-friendly tap-to-send path that needs no worker.
- Migrated the data layer from Google Sheets to Supabase, then removed the migration-era compatibility shims once they stopped earning their keep.
- Shipped a public demo on the same code: six invented tenants and two years of history generated relative to today and rebuilt nightly, behind three guards so the reset can never reach family data.
- Covered the rent math, access tiers, insights, and API routes with 61 test files.
Next.js 16
Claude API
Gemini API
Supabase
Zod
PWA
- Built household expense tracking with a catalogue of items and categories, monthly summaries, category breakdowns, and spending analytics.
- Implemented slip scanning that reads a photographed handwritten page into draft rows through a vision model, replacing manual retyping.
- Grounded the model in the household's own item catalogue so it returns familiar Hinglish spellings and links lines to existing items.
- Handled the day-first Indian date convention explicitly in the prompt, eliminating a whole class of month-day transposition errors.
- Modelled a photographed page as a multi-day document read from up to four photos, so lines keep their own dates and a page spanning a month boundary lands in both months.
- Built a mandatory review screen between model output and the database, with uncertainty flags and the slip's own total reconciled against parsed lines.
- Supported two vision providers behind one interface, preferring the more accurate paid reader when configured and falling back to a free tier.
- Documented the free tier's data-use terms in the setup guide and kept the feature fully optional, with manual entry designed to stand alone.
- Built insights that notice gaps — a missed bill, an item due for re-purchase — and put the action that resolves each one on the same row, opening a pre-filled entry rather than saving blind.
- Kept classification, summary math, and analytics in pure server-side modules with their own unit tests rather than computing in components.
Swift
SpriteKit
iOS
macOS
Game Design
- Built a casual pick-up-and-play game about the sea in SpriteKit, with iOS and macOS targets sharing a single game core.
- Implemented a pseudo-3D perspective sea sailed toward a horizon, getting depth and distance from a 2D framework without a 3D engine.
- Designed the whole mechanical vocabulary around one input — steering — including whirlpools that pull the ship and are fought by steering out.
- Tied difficulty to distance with waves that damage, slow, and stiffen the further out the player sails.
- Added a reveal on the horizon after sustained progress, motivating continued play through curiosity rather than a score counter.
- Instituted a written four-gate design filter every feature must pass before being built, with failing ideas parked in their own document.
- Kept nine living design documents and an append-only decision log recording the reasoning behind each choice, treating code-doc disagreement as a bug.
- Sequenced milestones feel-first and left shooting out of the core loop, despite the project's name, because it failed the design filter.
Swift
SwiftUI
Swift Package
Swift Testing
iOS
- Building a graph-first family tree app where relatives are added from the tree itself, with placeholder slots for missing parents, partners, and children.
- Rebuilt the layout as a pure-Swift engine in its own package that never imports SwiftUI, shrinking the root view from 1,640 lines to 93.
- Stored only parent-child and partner relationships and derived siblings, labels, and family groups, removing a whole class of inconsistent data.
- Fixed relatives landing beside strangers by deciding each row's order before any coordinates exist, keeping partners and co-parents together.
- Covered family rules and golden layouts with 80 package tests that run in about a second, without a simulator.
- Working from twelve design documents, a decision log, and an ordered plan; next up are saved data, editing with undo, and accessibility.
College projects
- Investigated paradoxes and mathematical properties within the quantum pigeonhole principle and pseudo-telepathy.
- Built a network with a Taskmaster node coordinating shared game state for multiple players.
- Designed controls to improve pilot comfort when using unfamiliar helicopter flight hardware.
- Built a bot that automates Duolingo German lessons, detects failures, and resets itself for continuous progress.
- Implemented a unique triangle-based RANSAC variation instead of the conventional L2 distance.
Skills
Languages
Java
Python
JavaScript
Backend
Spring Boot
REST APIs
Microservices
Data & Storage
PostgreSQL
MongoDB
MSSQL
InfluxDB
Systems & Tools
Docker
Grafana
Jenkins
Linux
Concepts
Distributed Systems
Event-Driven Architecture
TDD
Education
M.Tech · Embedded Systems
NIT Jaipur
June 2021
B.E · Electronics and Telecommunications
Jabalpur Engineering College
June 2018