Niranjan Sonawane — Backend Engineer

Full Stack Backend Engineer (Full-time) — Tynary

Feb 2026 — Present · Remote

Logistics routing, event-driven microservices and distributed background processing.

• Design a logistics routing engine for delivery optimisation.
• Architect event-driven microservices in Go and RabbitMQ for image-upscaling and background-removal workloads, orchestrated with Docker Compose.
• Configure NGINX reverse proxies, load balancing and CI/CD pipeline strategies for zero-downtime deployments on high-availability backend systems.
• Design secure JWT / OAuth2 authentication and RBAC authorisation with a scalable schema architecture.

• Cut average route computation time by 40% and reduced dispatcher workload by 2 hours a day.
• Enabled async processing and 3x horizontal scalability for image workloads.
• Built distributed background pipelines handling 500+ concurrent jobs at 99.9% uptime with fault-tolerant task recovery.

Technologies: Golang, RabbitMQ, Docker Compose, NGINX, CI/CD, JWT / OAuth2, RBAC

Django Developer (Part-time) — Phalmora

Jan 2026 — Feb 2026 · Remote

Django backend and REST APIs for a School Management System with a Flutter mobile app.

• Developed and maintained scalable backend services for a School Management System, supporting student, teacher and admin workflows.
• Designed and implemented RESTful APIs for seamless integration with a Flutter-based mobile application.
• Integrated authentication and role-based access control for secure multi-user access (admin, teacher, student).
• Collaborated closely with the Flutter team to optimise API performance and reduce latency.

• Enabled real-time data synchronisation between backend and mobile app across attendance, notifications and academic records.

Technologies: Django, REST APIs, Flutter integration, RBAC

Software Engineer Intern — Leadequator

Nov 2025 — Dec 2025 · Pune, Maharashtra · Remote

AI agents, data pipelines and VPS deployments for business intelligence.

• Developed intelligent AI agents to collect, process and analyse data for business intelligence and decision-making.
• Built data pipelines to extract insights from multiple sources, enabling automated reporting and analytics workflows.
• Designed backend services to support insight generation systems, improving data accessibility for business use cases.
• Configured servers, handled deployments and optimised system reliability for continuous data processing workloads.

• Deployed and managed production-ready services on VPS environments, ensuring high availability and performance.

Technologies: AI agents, Data pipelines, Backend services, VPS, Deployments

Backend Engineer Intern — Thinkdex Technology

Feb 2025 — May 2025 · Remote

Backend migration, cloud deployment and a recommendation system.

• Migrated a legacy Python codebase to a scalable Golang microservices architecture.
• Deployed containerised applications on AWS EC2 using Docker.
• Configured NGINX for load balancing.

• Improved system performance through the move to Go microservices.
• Developed a machine-learning recommendation system using behavioural analytics for personalised content.
• Reduced AWS costs by 25% through auto-scaling and load-balancing strategies.
• Refactored a legacy Python monolith into high-concurrency Golang microservices: +60% API throughput and -35% P95 latency.
• Automated CI/CD pipelines with Docker Compose, NGINX and AWS EC2, reducing deployment time by 50%.
• Engineered an async recommendation pipeline with decoupled task execution, cutting request critical-path latency by 45% for 10k+ daily active users.
• Reduced infrastructure cost by 25% through optimised container resource allocation and right-sized EC2 instances.

Technologies: Golang, Python, AWS EC2, Docker, NGINX, Machine learning

Head of AI/ML Domain (Volunteer, alongside studies) — Google Developer Groups (GDG) SKNCOE

Aug 2024 — Present · Pune, Maharashtra

Leading AI/ML learning for a student developer community.

• Lead AI/ML initiatives: workshops, research projects and mentorship for 100+ developers.
• Collaborate with cross-functional teams to organise AI hackathons and model-deployment sessions.

• Guided members in building LLM-powered applications, RAG pipelines and real-time ML-driven solutions.
• Led 8+ workshops and 3 hackathons on LLMs, RAG, vector databases and AI engineering.
• Mentored 100+ students on production ML deployment strategies.

Technologies: LLMs, RAG, Vector databases, LangChain, Python

B.E. in Computer Science — Savitribai Phule Pune University (SPPU)

Aug 2023 — May 2027 (expected) · Pune, Maharashtra

Bachelor of Engineering, CGPA 8.5 / 10.0.

• Computer Science coursework alongside hands-on backend and AI projects.

• CGPA 8.5 / 10.0.

Technologies: C++, Python, Go, SQL

Real-Time MCQ Assessment Platform

Live assessments for 1000+ concurrent users. A scalable real-time MCQ platform in Go with WebSocket connections, Redis caching for live leaderboards and an admin dashboard with live analytics.

Stack: Golang, WebSockets, Redis, MongoDB

• WebSocket server in Go built on goroutines for concurrent connections.
• Redis caching for session management and live leaderboard updates with sub-second latency.
• Live admin dashboard with real-time analytics on user responses and test progress.

A real-time platform supporting 1000+ concurrent users, with Redis-backed leaderboard updates at sub-second latency.

Source code

Connectree — Distributed Real-Time Messaging

10k+ concurrent WebSocket connections, under 10ms delivery. A low-latency messaging platform in Go: WebSocket connections, Redis Pub/Sub for horizontal scaling and in-memory matchmaking that keeps load off the database.

Stack: Golang, Redis, WebSockets, Gin

• Low-latency message delivery (<10ms) using Go goroutines and channels.
• Redis Pub/Sub for horizontal scaling across 5+ backend instances.
• In-memory matchmaking that reduced DB read load by 70% during peak traffic.
• Gin HTTP layer for connection upgrade and management endpoints.

Supports 10k+ concurrent WebSocket connections with sub-10ms delivery, scaled over 5+ instances, with 70% lower database reads at peak.

Source code

Perplexity-Style AI Search Engine

Research answers with real-time web data and citations. AI-powered search with real-time web scraping, citation tracking and source verification on a RAG architecture, using multiple LLMs.

Stack: Python, LangChain, RAG, Web scraping

• Real-time web scraping feeding a retrieval-augmented generation pipeline.
• Citation tracking so each claim links back to its source.
• Multiple LLM models integrated with source verification for deeper research.

A working search-engine backend that returns researched answers with tracked citations.

Source code

Multi-Agent RAG Platform

Parallel retrieval agents with citation traceability. A multi-agent retrieval-augmented generation platform: three agents search in parallel and every claim maps deterministically to its source.

Stack: Python, LangChain, FAISS, Multi-agent RAG

• Parallel semantic search across 3 retrieval agents.
• Deterministic citation mapping from retrieved passages to answer text.
• LangChain orchestration over FAISS vector indexes.

Answer traceability improved to 95% source-attribution accuracy.

Source code

Music Recommendation System

Content-based recommendations with PCA and cosine similarity. A content-based music recommender built with scikit-learn and served through Django.

Stack: Python, Scikit-learn, Django, Machine learning

• Content-based recommender using cosine similarity.
• PCA dimensionality reduction on the feature space.
• Django application exposing the recommender.

A complete content-based recommender (Sep 2024).

Source code

AI-Powered ARGO Oceanographic Data Explorer

Ask questions about ocean float data in plain English. A conversational AI system for querying ARGO float data with RAG and multimodal LLMs, backed by an ETL pipeline and a React dashboard with geospatial visualisations.

Stack: Python, Go, RAG, React.js, PostgreSQL, FAISS, ChromaDB, NetCDF

• ETL pipeline converting NetCDF oceanographic data into PostgreSQL and FAISS.
• RAG pipelines with multimodal LLMs for conversational querying.
• React.js dashboard with geospatial visualisations driven by natural-language queries.
• Go ingestion engine processing 2M+ oceanographic records from NetCDF datasets, indexed into ChromaDB with sub-200ms semantic retrieval latency.
• RAG pipeline with citation-based retrieval, reducing the hallucination rate by ~60% versus baseline LLM responses on domain-specific queries.

A deployed, working explorer with a live demo and a public backend repository: 2M+ records ingested, sub-200ms semantic retrieval and ~60% fewer hallucinations than a baseline LLM on domain queries.

Source code · Live demo

GDG SKNCOE — AI/ML Domain

Workshops, hackathons and mentorship for 100+ developers. Leading AI/ML initiatives at Google Developer Groups SKNCOE: workshops, research projects, hackathons and model-deployment sessions.

Stack: LLMs, RAG, Mentorship, Hackathons

• Workshops and research projects on AI/ML topics.
• AI-focused hackathons and model-deployment sessions with cross-functional teams.
• Guiding members to build LLM-powered applications, RAG pipelines and real-time ML solutions.

Ongoing: mentoring 100+ developers and growing the community’s AI/ML work.

Source code