Hi, my name is

SUNDARAM RAI

I build backend-heavy products with sharp interfaces: tunnels, secure vaults, RAG systems, payment flows, and internal tools that survive real users.

Selected Systems Download Resume

SYSTEMS THAT SHIP • BACKENDS WITH PRODUCT SENSE • INTERFACES FOR REAL WORKFLOWS •

01.
Where I've Built

The resume has the full inventory. This is the compressed version: the kind of systems I have been trusted to move in production contexts.

Associate Software Engineer

Homeville Group

Jan 2025 - May 2026

Loan workflows, document vaults, payment verification, AI-extracted data, and real-time internal systems across CRM, LOS, and LMS surfaces.

Web Developer Intern

Zidio Development

Jun 2024 - Sep 2024

Built a role-aware hiring platform with auth, search, pagination, application tracking, and candidate management APIs.

Software Developer Intern

Tata Steel Limited

Aug 2023 - Jan 2024

Turned manufacturing exports into plant-level OEE dashboards and forecasting pipelines for production planning.

02.
Selected Systems

Four builds that show the range better than a stack list can: networking, security, retrieval, and applied AI.

Vexlo

Problem: local services are hard to expose, inspect, and replay without depending on a hosted tunnel.

Build: a self-hosted Go tunnel with binary TCP transport, framed forwarding, request replay, SQLite persistence, WebSocket inspection, and ACME TLS.

Go SQLite WebSocket TCP REST API GitHub Actions

Cipheria

Problem: password managers fail if the server becomes the place where trust concentrates.

Build: client-side AES-256-GCM encryption, PBKDF2-SHA256 key derivation, Redis rate limits, and JWT rotation with reuse detection.

Next.js FastAPI PostgreSQL Alembic Redis AES-256-GCM JWT

Berkshire Hathaway Intelligence

Problem: decades of shareholder letters are dense, valuable, and slow to interrogate manually.

Build: a RAG chat system with PDF ingestion, MDocument chunking, pgvector retrieval, Mastra agent orchestration, memory, and streaming source attribution.

Next.js TypeScript Mastra PostgreSQL pgvector Mistral Embeddings Ollama

Juris AI

Frontend Backend

Problem: legal answers need current source grounding, not just confident model recall.

Build: an Angular and Node.js platform that extracts legal intent, retrieves live web context, and injects it into Gemini generation.

Angular Node.js MongoDB Gemini API JWT

03.
Engineering Range

I do not think in tool logos. I think in constraints: latency, trust boundaries, messy data, handoffs, and the point where a workflow breaks.

01 / Networked Tools

When the product is a protocol

Tunnels, framed transports, WebSocket inspection, replay systems, CLI-friendly flows, and the operational edges around them.

Go / TCP / SQLite / WebSocket / GitHub Actions

02 / Trust Boundaries

When the server should know less

Client-side encryption, token rotation, webhook verification, access scopes, document storage, and payment events that need to be correct.

FastAPI / Node.js / JWT / Redis / AWS S3 / MySQL

03 / Retrieval Systems

When answers need receipts

PDF ingestion, chunking, embeddings, pgvector retrieval, source attribution, query rewriting, and agent flows that stay grounded.

Next.js / TypeScript / PostgreSQL / pgvector / Mastra / Gemini

04 / Internal Software

When speed depends on workflow clarity

Dashboards, loan journeys, manufacturing metrics, role-aware portals, and interfaces that turn operational noise into decisions.

Angular / React / REST APIs / MongoDB / Firebase / Python