Skip to content

Available for new opportunities

Mahavir Vaishnav

I buildGenAI-powered solutions

I build scalable backend systems, enterprise applications, and GenAI-powered solutions — Node and FastAPI services, RAG pipelines, and MCP tooling that turn five-minute queries into ten-second ones, and prototypes into systems that hold up under real load.

View my workRésumé

Mumbai, IndiaRemote-friendly

profile.jsonlive
Role
Senior Software Developer
Focus
GenAI · RAG · MCP
Runtime
Node.js & TypeScript
Runtime
Python & FastAPI
Data
PostgreSQL · MongoDB · Vector DBs
Leads
6 engineers
Node.jsTypeScriptPythonFastAPINext.jsAWS
01About

Backend first, AI on top.

Node and TypeScript are the foundation; the GenAI, agent, and retrieval work sits on top of it. Same engineer, one more layer.

Senior Software Developer with 5+ years of experience building scalable backend systems, enterprise applications, distributed systems, and GenAI-powered solutions. Hands-on across Node.js, FastAPI, Python, TypeScript, LLMs, RAG, MCP, vector databases, embeddings, and semantic search — designing retrieval pipelines, API-driven architectures, event-driven systems, and high-volume data platforms.

Node.js and TypeScript have carried all five of those years — Express services, event-driven workflows, and the transactional cores these platforms run on. The last two added a GenAI layer on top of that foundation: LLM integrations, RAG assistants grounded in embeddings and semantic search over ChromaDB and Pinecone, and MCP tools that let an assistant compute, process data, and hand back a finished PDF or Excel file instead of a paragraph — all served from production FastAPI services. The track record is in the numbers: PostgreSQL queries from 5 minutes to under 10 seconds, 6 engineers led, and platforms serving 5,000+ merchants at 1,000+ orders an hour.

Mumbai, India · Remote-friendly

5+
Years building for production
1,000+
Orders processed per hour
5,000+
Merchants on platforms I built
6
Engineers led

Available for new opportunities— Open to remote roles worldwide

Get in touch
02Skills

Two runtimes, one system.

Node and TypeScript carry the transactional core; Python and FastAPI carry the AI and data layer. I design the seam between them — and everything either side needs to stay fast, observable, and cheap to run.

Node.js & TypeScript

5+ years in production

The foundation under every platform I have shipped — the transactional core the Python services sit next to, not replace.

  • Express.js REST services
  • Event-driven architecture
  • RabbitMQ async workflows
  • Socket.io real-time dashboards
  • Third-party integration layers
  • API design and authorization

Python & FastAPI

The AI and data layer

Async services that carry the LLM, retrieval, and document workloads, with Pydantic contracts at every boundary.

  • Async FastAPI microservices
  • Pydantic-validated contracts
  • LLM, RAG, and agent endpoints
  • Background workers for long jobs
  • Pandas data pipelines
  • Auto-generated OpenAPI docs
Node.jsTypeScriptExpress.jsPythonFastAPILLM AgentsRAGMCPpgvectorChromaDBPineconeOpenAIClaude APIReactNext.jsPostgreSQLMongoDBRedisRabbitMQSocket.ioAWSDocker

AI & LLM Engineering

  • LLM Integration
  • OpenAI APIs
  • Claude API
  • Prompt Engineering
  • Structured Outputs
  • Function / Tool Calling
  • Streaming Responses
  • Evals & Guardrails
  • Token & Cost Optimization

Agents & MCP

  • AI Agent Workflows
  • Multi-Step Orchestration
  • Tool-Calling Agents
  • Human-in-the-Loop Approval
  • MCP Servers & Clients
  • Claude Code

RAG & Vector Databases

  • RAG Pipelines
  • pgvector
  • ChromaDB
  • Pinecone
  • Embeddings
  • Chunking Strategies
  • Hybrid & Semantic Search
  • Re-ranking

Languages

  • TypeScript
  • JavaScript (ES6+)
  • Python
  • SQL

Backend

  • Node.js
  • Express.js
  • TypeScript APIs
  • FastAPI
  • Pydantic
  • Async Python
  • REST APIs
  • API Design
  • Auth & Authorization

Frontend

  • React.js
  • Next.js
  • Redux
  • HTML5
  • CSS3
  • Tailwind CSS

Databases

  • PostgreSQL
  • MongoDB
  • MySQL
  • Microsoft SQL Server
  • Indexing & Query Tuning

System Design

  • Software Architecture
  • Distributed Systems
  • Event-Driven Architecture
  • Microservices
  • Scalability
  • Performance Optimization

Messaging & Real-Time

  • RabbitMQ
  • Redis
  • Socket.io
  • WebSockets

Cloud & DevOps

  • AWS (Lambda, EC2, API Gateway, SES, S3, Route 53, Amplify)
  • Docker
  • CI/CD
  • Linux
  • Nginx
  • PM2
  • GitHub Actions

Data & Analytics

  • Pandas
  • NumPy
  • Data Pipelines
  • Analytics Reporting
  • GA4

Leadership & Process

  • Team Leadership
  • Technical Mentoring
  • Forward Deployed Engineering
  • Agile / Scrum
  • Jira
  • Confluence

Tools & Platforms

  • Git
  • GitHub
  • Sentry
  • Postman
  • Jest
  • Swagger

Integrations

  • Shopify
  • WooCommerce
  • Amazon SP API
  • Bloomreach
  • Zoho
  • Razorpay
  • PhonePe
  • Paytm
  • JusPay
03AI Engineering

What I build with LLMs.

Applied AI in production systems, not prototypes — agents that call real tools, retrieval grounded in embeddings and semantic search, and FastAPI services with evals, guardrails, and cost limits wired in from the start.

LLM-Powered Products

Production features built on OpenAI and Claude APIs — not demos. Schema-grounded prompts, structured outputs the rest of the system can trust, streaming responses, and fallbacks for when a model is slow or wrong.

  • Personalised AI workflows over ingested lead and customer data
  • Structured extraction with strict JSON schemas and validation
  • Streaming token delivery over SSE for sub-second perceived latency
  • Prompt versioning, evals, and regression checks before rollout
OpenAIClaude APIPydanticFastAPISSE

AI Agent Workflows

Multi-step agents that actually do the work: plan, call internal tools, check their output, and escalate to a human when confidence drops. Built with retries, idempotency, and a hard ceiling on spend.

  • Tool-calling agents wired to internal APIs and databases
  • Multi-step orchestration with retries, timeouts, and fallbacks
  • Human-in-the-loop approval gates on business-critical actions
  • Per-run token, latency, and cost tracing for every agent step
Tool CallingCeleryRedis

RAG & Vector Search

Retrieval pipelines that return the right chunk instead of a plausible one — deliberate chunking, hybrid search, and re-ranking, with citations back to the source document.

  • Embedding and ingestion pipelines over documents and structured records
  • Hybrid keyword + vector search with re-ranking for precision
  • pgvector alongside Postgres; ChromaDB and Pinecone for standalone collections
  • Answer grounding with citations and retrieval-quality evals
pgvectorChromaDBPineconeEmbeddingsPostgreSQL

MCP & Tooling

Model Context Protocol servers that expose internal systems to AI assistants safely — scoped, authenticated tools with auditable calls, so teams query live data from Claude instead of filing a ticket.

  • MCP tools for computation, data processing, and document generation
  • Automated personalised PDF and Excel deliverables from an agent run
  • Scoped, authenticated, audit-logged tool surfaces
  • Claude Code adoption and internal developer tooling
  • Model-agnostic abstractions so providers swap without a rewrite
MCPClaude CodeNode.jsTypeScriptPythonOAuth

Service Architecture

Node.js and Express own the transactional core; async FastAPI services sit beside them for the AI, data, and document workloads — one system split along a clean seam, each half deployed in Docker and scaled on its own curve.

  • Event-driven Node services owning the transactional path
  • Clean REST contracts between the Node and Python halves
  • Async FastAPI microservices with Pydantic-validated contracts
  • Background workers for long-running AI and data jobs
  • Dockerised deploys, health checks, and independent scaling
Node.jsTypeScriptFastAPIDockerOpenAPI

Data & Automation

Python and Pandas pipelines that take recurring analyst work off people's plates — large-scale reporting, reconciliation, and the unglamorous glue that keeps numbers trustworthy.

  • Large-scale reporting and operations automation in Python and Pandas
  • 80% of recurring manual operations removed at Nexsales
  • Scheduled pipelines with validation and failure alerting
  • AI-assisted summaries layered on top of deterministic data
PythonPandasNumPyPostgreSQLCron
04Experience

Where I've worked.

5+ years shipping production systems across enterprise GenAI, logistics, ed-tech, and consulting — most recently leading a team of 6 and building the LLM, RAG, and MCP layers on top of the Node and TypeScript platforms underneath.

Feb 2025 – May 2026

Nexsales Solution Pvt Ltd

Senior Software Developer

Andheri, Mumbai

  • 5 min → 10 s queries
  • 6 engineers led
  • 80% less manual ops
  • GenAI in production
  • Built a GenAI solution for a US enterprise client on Python, FastAPI, and MCP — ingesting lead and customer data from Bloomreach and email services, embedding it into ChromaDB and Pinecone, and serving it back as semantic retrieval behind personalised AI workflows.
  • Shipped a RAG-based AI assistant that answers in context rather than plausibly — LLM generation grounded in vector retrieval, embeddings, semantic search, and contextual prompting tuned against real client data.
  • Gave that assistant hands through MCP tools — computation, data processing, and automated generation of personalised PDF and Excel deliverables, so the answer arrives as a finished document instead of a paragraph.
  • Designed the Node.js and TypeScript microservices and REST APIs underneath — PostgreSQL, Redis, AWS, Docker, and CI/CD carrying distributed, event-driven workflows, authentication and authorisation, and high-volume data processing.
  • Took the platform's worst PostgreSQL queries from 5 minutes to under 10 seconds — profiling query plans, designing the right indexes, and refactoring the SQL behind enterprise reporting workloads.
  • Led 6 engineers across system design, architecture, feature delivery, and stakeholder collaboration, setting the technical direction with data teams and senior leadership.
  • Operated as a Forward Deployed Engineer with two US-based enterprise clients — gathering requirements first-hand, driving solution discussions, and shipping business-critical features against real deadlines.
  • Built Python and Pandas automation over recurring operational work, removing 80% of the team's manual effort.
  • Ran Agile delivery end to end through Jira and Confluence, making releases predictable and knowledge shared instead of tribal.

Jul 2023 – Jan 2025

Shipyaari

Full Stack Developer

Malad, Mumbai

  • 5,000+ merchants
  • 1,000+ orders/hour
  • 98% success rate
  • 60% faster APIs
  • Architected and scaled Blaze, a Node.js and TypeScript logistics platform serving 5,000+ merchants — sustaining 1,000+ orders per hour across 100+ concurrent users at a 98% order-processing success rate.
  • Re-engineered high-volume operations onto RabbitMQ-backed asynchronous, event-driven workflows, decoupling the critical path to lift API performance 60% and keep the system resilient under burst load.
  • Integrated OpenAI APIs into backend workflows to bring LLM-powered capabilities into the logistics pipeline, kept behind clean service contracts so the transactional path stayed unaffected.
  • Owned the core shipment engine: lifecycle management, rate calculation, courier selection, and tracking workflows spanning multiple logistics partners.
  • Integrated Shopify, WooCommerce, Amazon SP-API, and Zoho so merchant catalogues and orders stayed in sync across systems without manual reconciliation.
  • Integrated PhonePe, Paytm, Razorpay, and JusPay for secure transaction processing and streamlined financial operations.
  • Built real-time shipment visibility on Socket.io and WebSockets, and shipped it through Docker-based CI/CD — merchants and internal logistics teams got a live view instead of a stale one.

Jan 2022 – Apr 2023

Master-O

Software Development Engineer

Remote

  • Built scalable React.js, Next.js, and Node.js applications on MongoDB and AWS behind REST APIs — real-time educational games, assessment modules, and configurable multimedia learning platforms.
  • Integrated enterprise-grade HRMS APIs and automated backend workflows with cron jobs, Docker, and CI/CD, cutting recurring manual processes out of the loop.
  • Shipped the Learning Aid Card (LA Cards) platform — swipe-based navigation with rich multimedia content including images and video.
  • Designed a reusable, configurable component system that let non-engineers compose entirely new learning experiences from admin-managed content.

Aug 2020 – Dec 2021

Appectual

Software Development Engineer

Malad, Mumbai

  • Built responsive React.js and Next.js web applications delivering a consistent experience across desktop and mobile.
  • Built and maintained Node.js, Express.js, and MongoDB REST APIs powering business workflows and third-party integrations.
  • Worked alongside engineering and product to develop, test, and ship full-stack applications, contributing to performance work and feature delivery.
05Featured Work

Products I've built.

Founder-led products where I owned every decision — architecture, infrastructure, interface, and go-to-market.

PdfRaven

PDF SaaS, designed and shipped solo

An end-to-end PDF SaaS — redaction, editing, merging, compression, and format conversion — built on a FastAPI document-processing backend behind REST APIs, with a server-rendered Next.js frontend tuned for speed and SEO. The architecture separates document workflows, auth, payments, and processing pipelines so LLM-powered editing and intelligent automation drop in without a rewrite. Every paying user arrived organically through technical SEO, measured in GA4, with zero paid marketing.

  • 8 active Pro users
  • 100% organic acquisition
Next.jsFastAPIMongoDBTailwind CSSGA4
Founder & Software EngineerJun 2026 – Present

Trenold

AI conversational platform

An AI-powered conversational platform for content generation, question answering, and productivity workflows. Built on OpenAI APIs with real-time prompt-based interactions over a web interface, and a frontend/backend split deliberately kept model-agnostic so additional LLM providers plug in as first-class citizens.

ReactNode.jsOpenAI APIsCloud APIs
Founder & Software EngineerJan 2025 – Feb 2025

Education

Bachelor of Engineering · Atharva College of Engineering

Malad, Mumbai

Aug 2016 – Jul 2020
06Contact

Let's build something good.

I'm open to senior engineering roles, consulting, and interesting collaborations. The fastest way to reach me is email.

Email me
GitHubLinkedInRésumé