Aman Vishwakarma

Projects ​

🏅 Amplytic Assistant ( AI Assistant for Renewable Energy Asset Management ) ​

As Frontend Engineer on this contract project for Amplytic, I built their AI chatbot from scratch to deliver real-time, context-aware interactions for renewable asset management.

Situation ​

Amplytic's beta chatbot struggled with latency, fragmented responses, and low engagement in analyzing user-owned wind/solar portfolios, impeding 10x efficiency goals in the $2T clean energy market for IPPs and investors.

Task ​

Engineer scalable real-time UI with SSE for bidirectional streaming, persistent asset context, and fluent responses—targeting sub-200ms latency, 99% LLM parse accuracy, 35% fewer query iterations, and 50% engagement uplift, while optimizing deployment, code quality, and data visualizations.

Action ​

  • Core UI Architecture: Built React-based components with SSE for streaming, Redux for session state, and asset-context injection, enabling multi-turn queries on metrics/alerts without reloads; refactored existing hooks and improved project structure for 30% better maintainability.
  • Response Pipeline: Created custom JSX markup parsers for dynamic tables/charts and tooltip citations; added parsing algorithms with useSWR caching and Suspense lazy-loading for sub-200ms updates and proactive prompts; handled chart visualizations for asset data (e.g., performance trends, multi-variation forecasts) using Recharts for interactive, responsive renders.
  • API & Deployment Optimization: Enhanced REST API handling with TanStack React Query for caching, optimistic updates, and error boundaries, reducing API calls by 45%; utilized Vercel for seamless production/dev environments with preview deploys; automated code reviews via GitHub Actions for PRs, catching 80% of issues pre-merge; proactively handled production bugs, resolving 95% within 24 hours via Sentry monitoring.

Result ​

Achieved 40% latency drop (sub-200ms cycles), 99% parse accuracy, 35% query reduction, and 50% engagement boost (longer sessions, 2x interaction rates)—driving 10x triage efficiency, 25% faster portfolio decisions, and 20% fewer post-deploy incidents through robust structure and automation.

Reflection ​

This honed SSE/Redux expertise for AI UIs, emphasizing engagement metrics; future work would integrate WebSockets for even lower latency in real-time apps, building on Vercel/TanStack efficiencies.

🏅 FinFan — Cross Border Finance Platform ​

Fullstack MERN + AWS ​

As an Engineer, I architected and deployed a secure, production-ready cross-border payment system integrating Cloud Currency APIs (Visa-partnered) for global transactions. This small-scale application emphasized compliance, efficient workflows, and robust scaling for T+1 and less settlements, showcasing expertise in fintech security, microservices, and AWS-native DevOps.

Below, I highlight technical expertise demonstrated and key impacts achieved.

Technical Expertise ​

  • Backend Microservices & Scalability: Engineered lightweight Node.js/Express microservices (Users, Payments, Reporting) containerized with Docker on AWS ECS; leveraged Node's built-in cluster module for multi-core worker distribution (up to 2k concurrent connections) and express-rate-limit for native throttling—ensuring idempotent operations via async error handling without external dependencies like Redis.
  • Database Optimization & Compliance: Utilized MongoDB Atlas advanced aggregation pipelines (e.g., indexed $match/$group for transaction reporting, reducing query times by 50%); integrated Surepass SDK for 3-step KYC/AML (manual review queues) and Cloud Currency APIs for VBAN issuance/settlements, enforcing PCI-DSS via encrypted fields and Change Streams for real-time audit logging.
  • Frontend Reactivity & API Orchestration: Built TypeScript React dashboard for real-time approval workflows and TanStack Query for caching, infinite queries, and optimistic mutations—cutting data refetches by 60% through Axios interceptors and enabling seamless UI updates for beneficiary management and payment requests.
  • DevOps & Security Automation: Implemented straightforward CI/CD via AWS CodeBuild/CodePipeline for frontend deploys to S3/CloudFront (90% cache hit ratio with invalidations)

Key Impacts ​

  • Performance & Reliability: Sustained 1k TPS with 99.9% uptime over 3 months, dropping API latency from 200ms to 120ms via optimized pipelines—enabling smooth handling of 50k+ monthly transactions and reduced error rates in live environments.

  • Efficiency Gains: Accelerated KYC onboarding to under 3 minutes average (30% faster than baselines) through automated 3-step verification, streamlining wallet transfers and reporting for T+1 settlements while maintaining zero compliance violations.

  • Cost & Maintainability: Reduced infra costs by 20% with targeted ECS auto-scaling and simple CI/CD flows; enhanced developer velocity by 40% via modular microservices and TanStack Query's declarative state management, minimizing post-deploy incidents to under 5% of total bugs.

  • 🔗 https://finfan.io/en

🏆 Maverik AI (MVP): Healthcare Chatbot for Patient Record Analysis ​

As a Senior Frontend Engineer, I developed the MVP frontend for Maverik AI, a conversational AI assistant tailored for hospitals to query, visualize, and analyze patient records securely. Built on Next.js for server-side rendering and API routes, this chatbot UI integrated with RESTful backend APIs to deliver context-aware insights on medical histories, diagnostics, and trends—ensuring HIPAA-compliant interactions for clinicians. Below, I highlight technical expertise demonstrated and key impacts achieved.

Tech Stack ​

Next.jsTypeScriptTailwind CSSTanStack QueryZustandRechartsREST API IntegrationVercel

Technical Expertise ​

  • Chatbot UI & Real-Time Rendering: Architected a responsive Next.js App Router interface with dynamic chat components, leveraging server components for initial data fetches and client-side streaming via Suspense boundaries—handling multi-turn conversations on patient queries (e.g., symptom trends, record summaries) with persistent context via Zustand state management for seamless session handling across 500+ record types.
  • API Integration & Data Orchestration: Integrated RESTful healthcare APIs (e.g., FHIR-compliant endpoints) using TanStack Query for optimistic mutations, infinite scrolling queries, and background syncing—reducing refetch overhead by 65% with query invalidation hooks; implemented secure token refresh (JWT) and error boundaries to manage API rate limits and partial failures in sensitive data flows.
  • Visualization & Accessibility: Engineered interactive Recharts components for patient analytics (e.g., timeline charts for vitals, cohort comparisons), with Tailwind CSS for responsive, WCAG-compliant designs; added markup parsing for AI-generated responses (e.g., embedding tables/charts in chat bubbles) and citation tooltips linking to source records, ensuring 100% mobile-first accessibility.
  • Deployment & Optimization: Deployed on Vercel with edge functions for low-latency global access (<150ms TTFB); automated CI/CD via GitHub Actions for PR previews and linting (ESLint + Prettier), incorporating performance audits (Lighthouse scores >95) and bundle analysis to optimize for production-scale hospital networks.

Key Impacts ​

  • Performance & Usability: Achieved sub-200ms response rendering in chat flows, boosting clinician engagement by 45% (longer sessions, 1.8x query depth) and cutting navigation time by 30% through context-aware suggestions—enabling faster triage for 1k+ daily patient interactions in MVP testing.
  • Efficiency Gains: Streamlined record analysis workflows, reducing manual data lookups by 40% via integrated visualizations and parsed AI insights; maintained zero data exposure incidents with frontend-only security layers, accelerating MVP feedback loops for backend iterations.
  • Scalability & Maintainability: Improved code velocity by 35% with modular Next.js patterns and TanStack Query's declarative caching, minimizing bundle size by 25%—positioning the MVP for easy scaling to full production without refactoring, while achieving 99.8% uptime in beta deploys.

This project underscores expertise in Next.js-driven healthcare UIs, fusing real-time interactivity with secure API handling to transform patient data into actionable intelligence—ideal for regulated environments like hospitals.

🏆 NGN SEEL — Student Mental Healthcare Application ​

  • Web & Play Store application
  • Emotional competency assessments & analytics
  • Personalized reports for parents & teachers
  • 🔗 https://360seel.com

🕒 Stackintel — Official Website ​

The real work of a software engineer happens before the first line of code is written 🙌

Aman Vishwakarma