Software Engineer · Pune, India

Sahas Nagar

Backend & Full-Stack Systems · Applied AI Engineering

I build reliable software systems end to end — REST APIs and data models, the applications on top, and the AI and cloud plumbing in between. Currently a founding engineer at OBLIQ-in, building a compliance platform; previously shipped document-intelligence, RAG, and booking systems.

View selected work Resume (PDF) GitHub LinkedIn

B.Tech Computer Engineering, MIT Academy of Engineering · 2027 sahas.p.nagar@gmail.com

How the pieces fit Fig. 01

SIH 2025
Smart India Hackathon finalist

#127
Amazon ML Challenge 2025 · global rank

3★
CodeChef rating

315+
DSA problems solved

About

One engineer, every layer of the system

I build software that ships — schema to deployed service, and increasingly the AI layer in between.

I'm a Computer Engineering student at MIT Academy of Engineering (graduating 2027) who has spent the last two years working on real systems rather than tutorials: a compliance platform at OBLIQ-in, document-intelligence pipelines, retrieval-augmented AI services, and a hostel-booking product taken from API design through to deployment.

My default stack is Java/Spring Boot and Python/FastAPI on the backend, React or Next.js on the front, with MySQL or PostgreSQL underneath and Docker plus a cloud platform to ship it. When a workflow deserves it, I add OCR, embeddings, or an agent pipeline — always as a component that solves a problem, never as decoration.

I care about the parts that outlive the demo: clear API contracts, migrations that don't lose data, logs you can actually trace, and deployments that survive contact with real users.

Own the whole stack

From schema design to the deploy pipeline — I want to understand, and be accountable for, every layer a system touches.

Contracts over cleverness

Clear REST contracts, explicit data models, and predictable error handling. Novelty belongs in the problem, not the plumbing.

AI as a component

OCR, retrieval, and LLM workflows earn their place when they remove real manual work — not because they're fashionable.

Ship, then harden

Working software under real constraints beats a polished spec. Deploy early, observe, then make it reliable.

Experience

Where I've built things

Three roles, one throughline: backend and full-stack delivery — APIs, data models, application workflows, and the deployments that keep them running.

Aug 2026 — Present Pune, India Current

Founding Engineer

OBLIQ-in

Contributing to the development of a production-oriented compliance platform across backend services, application workflows, database operations, and deployment.

Developing and integrating application features with a focus on maintainable backend architecture, API workflows, and reliable data processing; working on debugging, Git-based collaboration, containerization, and deployment workflows.

REST APIs SQL Docker Git

Apr 2026 — Jul 2026 Hybrid

Trainee, AI Full Stack Developer

Glimmora International

Developed backend and full-stack features using Java/Spring Boot and modern web technologies, integrating REST APIs and application authentication.

Worked on database-driven application workflows, debugging, Git-based development, and application-service deployment; contributed to document-processing modules and frontend–backend integration.

Java Spring Boot REST APIs Authentication SQL Git

Jun 2025 — Aug 2025 Remote

Full Stack Development Intern

Campus Credentials

Developed Spring Boot backend services with Spring Security, JWT authentication, and REST APIs for a hostel-booking platform.

Implemented APIs for property listings, booking workflows, search, filtering, and MySQL database operations; collaborated with a four-member team using Git and deployed the application on Render.

Spring Boot Spring Security JWT REST APIs MySQL Git Render

Selected work

Projects with real engineering behind them

Six flagship systems — retrieval platforms, document intelligence, booking infrastructure, and multi-agent AI — plus a wider directory of shipped work. Filter by domain, or open the full directory.

01 · 2026

REDROB

AI knowledge platform

A RAG knowledge platform with workspace-scoped auth, asynchronous document ingestion into pgvector, and LangGraph research workflows that answer questions with citations.

FastAPI PostgreSQL + pgvector Redis Celery LangGraph

Repository

Problem

Teams accumulate documents but can't query them together — ad-hoc chatbots lack workspace boundaries, provenance, and a real ingestion pipeline.

Engineering

Uploads land in S3-compatible storage; a Celery worker parses, deduplicates, chunks and embeds into PostgreSQL + pgvector. A LangGraph workflow retrieves under strict similarity thresholds and returns structured answers with citations, with request IDs and task IDs threaded through the logs.

02 · 2025 – 2026

DocuMitra

Document intelligence

An AI document-processing system that extracts text, classifies by department with confidence scores, summarizes, highlights, and routes documents automatically.

Python OCR NLP scikit-learn React n8n

Repository

Problem

Mixed-format documents have to be read, sorted and forwarded to the right team by hand — slow, inconsistent, and un-auditable at volume.

Engineering

An extraction pipeline (PyPDF2, python-docx, openpyxl, OCR) feeds a Naive Bayes classifier scoring 12 departments with keyword evidence; structured fields, summaries and highlighted copies are generated and routed downstream. Team project of six — I led the automation and processing pipeline.

03 · 2025

StayNest

Full-stack booking platform

A PG/hostel booking platform with Spring Boot REST APIs, JWT authentication with role-based access, property search and booking workflows, and a React frontend — deployed on Render.

Spring Boot Java 17 Spring Security JWT MySQL React

Repository Deployment repo

Problem

Students hunt for PG accommodation across fragmented listings; owners have no lightweight system for bookings, check-ins, and tenants.

Engineering

Spring Boot 3 backend with Spring Security, JWT auth, Spring Data JPA over MySQL and mail services; React 19 frontend with map integration and image hosting. Built with a four-person team on Git workflows during the Campus Credentials internship.

04 · 2026

MonArch

Multi-agent AI platform

A multi-agent orchestration platform: LangGraph agents for planning, research, hybrid RAG and vision, wrapped in guardrails, rate limiting, and an evaluation harness.

FastAPI LangGraph FAISS + BM25 Groq Vision Docker AWS

Repository

Problem

Single-shot LLM calls fail on complex tasks — no routing, no grounding in documents, and no protection against injection or hallucinated output.

Engineering

A FastAPI entrypoint fronts a sliding-window rate limiter and PII/injection guardrails; a LangGraph orchestrator routes to planner, researcher, hybrid RAG (FAISS + BM25) and vision agents, with a reflection critic retrying failed passes and an output-faithfulness check at the boundary.

05 · 2026

OrbitPMS

Property management system

Live demo

A hotel PMS covering rooms, reservations, check-in/out and invoicing — FastAPI + PostgreSQL with a React 19 frontend, JWT roles, and database-level double-booking protection.

FastAPI PostgreSQL SQLAlchemy React 19 JWT

Repository Live demo

Problem

Small hotels run bookings, front-desk state and invoicing across spreadsheets, with double bookings and no audit trail.

Engineering

Access/refresh JWT tokens with three roles (admin, receptionist, staff); room status is a validated state machine; bookings are protected twice — an application-level overlap check and a PostgreSQL GiST exclusion constraint.

06 · 2025

BhashaSetu

Offline AI translation

A fully offline OCR and translation system for Nepali and Sinhala literature — PaddleOCR, NLLB-200 and FastText running locally behind a Flask API.

Python Flask PaddleOCR NLLB-200 FastText

Repository

Problem

Translation tools assume connectivity and cloud GPUs; heritage and low-connectivity contexts need everything to run on the reader's own machine.

Engineering

A Flask backend orchestrates PaddleOCR for extraction, FastText for language detection and NLLB-200 for offline translation, with local summarization; an Electron-wrapped React frontend consumes the API and models run entirely on-device.

No projects match that search or filter.

More work

Smaller systems, hackathon builds, and experiments — including Sampark AI, ResolveX, Snapora, SpamShield, TaskSphere, and AWS automation tooling.

Open the full project directory

Technical skills

The tools I use to build

Grouped by where they sit in a system. Every item here appears in a resume or a shipped project — nothing is listed for the sake of it.

Languages

Daily drivers across backend, scripts, and frontends.

Java
Python
JavaScript
TypeScript
SQL
HTML/CSS

Backend & APIs

Service design, auth, and data access.

Spring Boot
Spring Security
Spring Data JPA
FastAPI
Flask
Django
REST API design
JWT

Frontend

Typed interfaces that talk to real APIs.

React
Next.js
TypeScript
Tailwind CSS

Databases

Relational modelling first; vectors when needed.

MySQL
PostgreSQL
pgvector
Oracle
Redis

Applied AI

Retrieval, classification, and agent workflows that ship.

RAG
LangGraph
LangChain
OCR
NLP
scikit-learn
Embeddings
LLM APIs

Cloud & DevOps

From local containers to production deploys.

Docker
AWS (EC2, S3, IAM, VPC)
Boto3
Git
Linux
CI/CD
Render
Vercel

Achievements

Competitions & recognition

National hackathons, a global ML competition, and consistent competitive programming.

Smart India Hackathon 2025 — Finalist

Selected as a national finalist in India's largest hackathon.

Amazon ML Challenge 2025 — Global Rank #127

Placed 127th worldwide among thousands of participants.

Atos Srijan Hackathon 2025 — Special Recognition

Awarded special recognition for the project submission.

CodeChef 3★

Achieved a 3-star rating through monthly contests and practice.

315+ DSA problems

Consistent problem-solving across platforms. Codolio profile.

Super 30 Scholar — MIT Academy of Engineering

Selected for the institute's merit scholarship programme.

Open source

Public GitHub activity

Recently updated non-fork repositories, loaded live from the GitHub API. Forks and experiments are excluded from this list.

—

Public repos (non-fork)

—

Stars earned

—

Languages in use

Loading repositories…

View full GitHub profile

Contact

Let's build something useful together.

I'm open to software engineering roles — backend, full-stack, or applied AI — and to interesting engineering problems in general. The fastest way to reach me is email.

Email sahas.p.nagar@gmail.com GitHub github.com/Sahas2711 LinkedIn linkedin.com/in/sahas-nagar Coding profile codolio.com/profile/Elite65

Open to opportunities

RolesSoftware / Backend / Full-Stack Engineer

Also open toApplied AI engineering

LocationPune, India · Remote-friendly

EducationB.Tech Computer Engineering, 2027

TimezoneIST (UTC+5:30)