Backend development & AI automation
Backend systems.
Practical AI.
I'm Akash. I build backend services, AI agents, and tools that help teams ship. My work connects reliable engineering with automation that makes a measurable difference.
Currently
Backend Engineer at Medable
Remote · May 2026 - present
- 2
- Production AI agents adopted organization-wide
- ~90%
- Faster notification delivery 5 days to under half a day
- 30+
- Client implementations delivered at Medable
02 / EXPERIENCE
Work that shipped.
From implementation to ownership.
Building production AI agents, notification infrastructure, and internal developer tools for an enterprise clinical trial platform.
- Independently initiated and shipped two production AI agents for custom backend development and notification troubleshooting, now adopted organization-wide.
- Built an orchestration agent that reduced notification delivery from 5 days to under half a day per client implementation, approximately 90% faster.
- Co-built Jira webhook integration to trigger root-cause analysis and fix merge requests directly from bug tickets, cutting debugging turnaround by approximately 90%.
How the systems work
From requirements to delivery
The orchestrator routes standard notification requirements to no-code configuration and custom requirements to code generation. Custom changes produce a merge request and go through approval before the agent deploys and configures them.
This lets project managers deliver standard notifications without developer involvement, while preserving review for custom code.
Tools that work under your own identity
Built a TypeScript Model Context Protocol (MCP) server with 21+ tools for platform configuration. Own both internal MCP servers and migrated them from shared deployment credentials to per-user authentication.
Configuration over repeated custom code
Introduced configurable, event-driven notification triggers, moving a category of custom backend logic into no-code workflows. The tooling also helps new engineers contribute from day one.
- Node.js
- JavaScript
- TypeScript
- MCP
- AI orchestration
Developed the backend foundations for reusable notification workflows, AI-assisted operations, and automated delivery.
- Built a TypeScript MCP server with 33+ tools so AI agents could query logs, inspect notification pipelines, and manage application records.
- Re-architected the JavaScript notification framework into reusable, configuration-driven components, eliminating duplicated business logic across implementations.
- Delivered notification workflows and custom backend functionality for 30+ client implementations, supporting GSK, AstraZeneca, AbbVie, and Novo Nordisk.
Delivery and testing improvements
Built GitLab CI/CD pipelines that automatically publish backend modules to a shared component library on merge to main, making them available across client implementations. Improved Jest test reporting to make failures easier to diagnose.
- JavaScript
- TypeScript
- MongoDB
- GitLab CI/CD
- Jest
- Developed FastAPI REST APIs for file uploads, authentication, and chat features.
- Built LangChain agents for stock and options queries, using Milvus for semantic search over user documents.
- Built reusable React components and complex data tables with Material UI, reducing subsequent UI development time by approximately 25%.
- Python
- FastAPI
- LangChain
- Milvus
- React
Personal projects in real-time infrastructure, AI retrieval, and distributed messaging. Each one explores a different engineering problem.
01 / EVENT INFRASTRUCTURE
ProducersGraphQL gatewayRedis Pub/SubRedis Streams
WebSocket clients
TypeScript · Backend infrastructure
Realtime Subscription Gateway
A multi-tenant event gateway that delivers live updates over GraphQL subscriptions and WebSockets, with Redis for fan-out and optional durable replay.
Engineering details
- Redis Streams supports replay from a sequence cursor, while Pub/Sub handles live event fan-out.
- Per-subscriber queues handle backpressure; token-bucket rate limits apply per tenant and topic.
- Tenant-scoped keys, Firebase JWT authentication, health checks, and Prometheus-style metrics address operational concerns.
- Node.js
- GraphQL
- Redis
- WebSocket
- Firebase Auth
View source for Realtime Subscription Gateway↗
02 / RETRIEVAL & REASONING
A legal questionKeyword + vector searchTool orchestrationReferenced answer
Python · AI agent system
AI Legal Help
An AI assistant for exploring the Indian Constitution and Indian Penal Code, combining document retrieval with multi-step tool orchestration.
Engineering details
- Uses LangGraph and a ReAct architecture to coordinate keyword extraction, vector search, summarization, and reasoning.
- Searches Constitution articles and IPC sections, cross-references retrieved material, and composes answers with references.
- Built with a FastAPI backend, Milvus vector database, and a React frontend.
- FastAPI
- LangChain
- LangGraph
- Milvus
- React
View source for AI Legal Help↗
03 / REAL-TIME MESSAGING
A shared roomNode.js + RedisMessagesPresence
Connected clients
TypeScript · Full stack
Scalable Chat App
A real-time chat application with shareable rooms, media sharing, and live presence. Guests can join a room using its ID and passcode without creating an account.
Engineering details
- Uses an asynchronous Node.js backend with Redis and Kafka for messaging infrastructure.
- Supports text, image, and video messages, plus online presence and typing indicators.
- Combines Next.js and Express with Prisma and SQL, with Google OAuth for account authentication.
- Node.js
- Next.js
- Redis
- Kafka
- Prisma
View source for Scalable Chat App↗
04 / TOOLKIT
What I work with.
Languages & backend
JavaScript, TypeScript, Python, Node.js, Express, FastAPI, GraphQL
Data & messaging
MongoDB, SQL, Redis, Kafka, Milvus, Prisma
AI & automation
Google Agent Development Kit (ADK), Model Context Protocol (MCP), LangChain, LangGraph, AI agent orchestration
Delivery & testing
GitLab CI/CD, Jest, Git, Jira
Frontend
React, Next.js, Material UI
Development tools
Cursor, Claude Code, Codex
Education2022 - 2026
B.Tech. in Computer Science
Software Product Engineering
Kalvium Undergraduate Program
Lovely Professional University