Senior Full-Stack Engineer · Open Source Maintainer · AI-Native Builder
Austin, TX · koushiksaha.dev · LinkedIn
Over a million people use software I've built every month. I've spent the last six years as a full-stack engineer working on problems that span the whole stack — micro-frontend architecture at 1M+ MAU scale, real-time data pipelines in Go, open-source design systems, and more recently AI-native products built on top of Claude APIs.
I care about building things that actually ship, hold up in production, and make the engineers around me faster. I'm a solo founder at Freedom Shopping LLC, an open-source maintainer, and a published researcher in AI/healthcare — not because those things look good on paper, but because they're the kind of work I find genuinely interesting.
The problem: A major US energy company needed its customer-facing web platform modernized — at 1M+ monthly active users, with zero downtime tolerance and three independent engineering teams who couldn't keep blocking each other on releases.
What I did: Picked Module Federation over a shared monorepo so all three teams could ship independently without coordinating deployments. Built a React 18 + TypeScript PWA that scored 95+ on Lighthouse. Added Go backend services for real-time energy monitoring data, handling high-throughput, low-latency feed ingestion using Go's concurrency model. Rebuilt the GitHub Actions CI/CD pipeline from scratch — releases went from 45 minutes to 8.
Outcomes: $2M+ ARR contribution, 30% cloud cost reduction (~$180K/year saved), 40% faster UI development via a shared Storybook component library, 82% faster release cycles.
A React + TypeScript + Storybook design system I built at Powerley and open-sourced. It handles the things design systems usually get wrong: WCAG 2.1 compliance baked in from the start, semantic versioning, and documentation that engineers actually read.
500+ weekly npm downloads · Used by 12+ companies · Single maintainer (me)
I manage the GitHub issues, community PRs, changelog, and semantic versioning myself. Building something other engineers trust enough to pull into their production codebases is a different kind of challenge than building for your own team.
The problem: Client teams were bottlenecked on frontend developers for every site update and new launch.
What I built: A WordPress-like system where the entire frontend — pages, navigation, menus, inputs, buttons, color schemes — is defined by a JSON payload. No frontend engineer required for maintenance or new launches. A backend developer or project manager can deploy a new site by providing JSON. One codebase, infinite configurations.
Impact: A single PM could launch 1,000+ websites simultaneously with all their changes. Zero frontend engineering time for ongoing maintenance. This system also helped close $500K+ in enterprise deals alongside a React SVG modifier library I built for physical security design teams — you could drag and drop entire building layouts and wire them to real-life sensor data.
An AI-native coaching PWA built entirely from scratch — Claude API, React, TypeScript, Next.js. Multi-turn prompt chains, LLM guardrails, and fallback logic to keep the experience reliable when model behavior is non-deterministic.
This is where I got serious about what it actually means to build a production product on top of an LLM, not just call an API and hope for the best.
A full-stack SaaS platform for academic publishing built on Next.js, TypeScript, PostgreSQL, and Prisma. It manages 12 journals with a custom 4-reviewer auto-publish workflow, ORCID login, Stripe/PayPal payment integration, peer review automation, PWA support, and a full CI/CD pipeline.
The interesting engineering challenge here was the reviewer assignment logic — building a workflow engine that could route submissions, track review state across multiple parties, and auto-publish without manual intervention.
A retail management platform covering daily/monthly sales reporting, inventory, automated payroll processing, tax exports, reconciliation, time tracking, and anomaly detection — across 20+ operational modules.
Multi-tenant from the ground up: RBAC permissions management, 2FA security, and data isolation across multiple companies and stores. The anomaly detection layer uses rule-based logic to flag unusual patterns in sales and inventory data before they become problems.
A full-stack real-time sign language translation service built to bridge communication for the hearing-impaired. The system uses MediaPipe for hand tracking, a Python/Flask backend for gesture recognition via computer vision, and WebRTC for live video chat integration — all tied together with a React and Next.js frontend.
The hard part was making real-time gesture classification reliable enough for live video at acceptable latency. The pipeline runs gesture recognition frame-by-frame via MediaPipe, classifies hand poses, and renders translated output in the video chat stream in near real-time. Deployed via Docker and Kubernetes with a CI/CD pipeline including security scanning (Trivy, CodeQL).
Rebuilt the mobile checkout flow for Bangladesh's largest B2C e-commerce platforms (moveon.com.bd, ali2bd.com) using React and Optimistic UI. Checkout conversion improved 35%, cart abandonment dropped 22%, Lighthouse score went from ~65 to 90+. Scaled the Laravel-based product search API 3x with no downtime during Black Friday traffic surges.
I got into AI/healthcare research because the problems are genuinely hard and the stakes are real. All published or in review in 2026:
Integrating Artificial Intelligence with Global Genomic Resources: A Narrative Review of Implications for Precision Medicine Dovepress · August 17, 2026 A narrative review of 54 studies synthesizing how AI integrates with global genomic resources across five application areas: genomic data integration, variant and disease association detection, disease susceptibility and risk prediction, treatment response prediction, and clinical decision support. Published open access.
Big Data in Cancer Genomics: Computational Foundations and Emerging Pathways for Precision Oncology Wiley (Computational and Systems Oncology) · July 9, 2026 · DOI: 10.1002/cso2.70020 Explores how computational foundations translate massive genomic datasets into actionable precision medicine insights. Accepted, in production.
Artificial Intelligence Applications for Hospital Resource Allocation in Emergency Preparedness Wiley (Advances in Public Health) · May 29, 2026 · DOI: 10.1155/adph/7200757 Examines how AI can optimize hospital resources — beds, staff, medical equipment — during crisis scenarios to strengthen healthcare resilience. Published, open access.
AI-Powered Predictive Analytics in Cancer Prevention Policy: A Systematic Review of Current Evidence and Future Directions Wiley (Health Science Reports) · Under Review A PRISMA-compliant systematic review across six major databases (2010–2026) examining AI-driven predictive models for cancer risk, screening optimization, resource allocation, and population-level forecasting. Findings show AI models outperforming traditional epidemiological frameworks by integrating clinical, behavioral, environmental, and social data.
Integrating Artificial Intelligence with a Global Atlas of Human Genetic Variation: Implications for Precision Medicine Journal of Multidisciplinary Healthcare · Under Review Co-author. Examines the intersection of AI and global genetic variation datasets and the implications for more inclusive, individualized clinical care.
Frontend React 18, Next.js, TypeScript, Module Federation, Storybook,
Tailwind CSS, Radix UI, shadcn/ui, Redux Toolkit, Jotai, Zustand,
TanStack Query, Webpack, Vite, Playwright, Cypress
Backend Node.js, Express, Go, GraphQL, REST, OpenAPI, Django, Laravel,
Prisma, PostgreSQL, MongoDB, Redis, ElasticSearch
AI/ML Claude API, multi-turn prompt chains, LLM guardrails,
Python/Flask, MediaPipe, WebRTC, computer vision
Cloud AWS (S3, Lambda, CloudFront, EC2), Docker, Kubernetes,
GitHub Actions, CI/CD pipeline design, Trivy, CodeQL
Certs AWS Solutions Architect – Associate (2025)
Google Cloud Professional Architect (2024)