ND
Narayan DewasiAI & Backend Engineer • Full Stack & UI Systems
AI & Backend Systems • Full Stack • 3+ Years Experience

Architecting fast, resilient backends, fluid UI & production AI.

Specialized In:AI Based Webapp Development

I'm Narayan Dewasi, an AI & Backend Engineer with deep Full Stack and UI expertise. Specializing in autonomous multi-agent pipelines (LangGraph), ridiculously fast Django & MERN applications, and clean, intuitive user interfaces. Known for first-principles problem solving and shipping reliable, end-to-end software.

Available for Full-Time & Freelance
3+ Years Battle-Tested
Pune, India
View Career Journey
Download Resume
Narayan Dewasi - AI & Backend Development
Django Experties60% Fast

Ridiculously fast Django APIs

Query optimization • Redis caching

AI Architecture ExpertiesLangGraph

Stateful multi-agent workflows

Dynamic RAG • Tool routing

Frontend & MERN ExpertiesFull Stack

Next.js App Router & UI Systems

WebSockets • Responsive layouts

Career Tree & Progression

The Engineering Journey

Rooted in core web engineering, branching into full-stack and distributed backends, and flourishing into production-grade autonomous AI systems.

The Tree of Growth

Milestones Along the Tree

Hover or follow the tree stages from root foundations to canopy AI.

Voxket.aiApex

Fullstack / Django & AI Architecture Developer

Jan 2026 – Oct 2026 • Remote

Slashash TechTrunk

Full Stack / Node.js Developer

Dec 2022 – Jan 2025 • In-Office

SharmaStoreRoots

Frontend Developer

Apr 2022 – Jun 2022 • Remote

CanopyVoxket.ai
TrunkSlashash
RootsSharmaStore
Canopy • Tree ApexRemote

Voxket.ai

Fullstack / Django & AI Architecture Developer

Jan 2026 – Oct 2026Remote

“Autonomous Agent Systems, LLM Fine-Tuning & High-Throughput APIs”

Pharma-Solver

FastAPI • LLM Fine-Tuning • Next.js
  • •Engineered an AI platform around a pharmaceutical fine-tuned LLM that flags flawed scientific processes and infers next actions.
  • •Cut hallucination rates by 40% and reduced inference latency by 70% through optimized token streaming and response caching.
  • •Built a distributed task-graph UI enabling researchers to decompose complex experiments into reviewable checkpoints.

LangGraph Multi-Agent Orchestration

LangGraph • Stateful Graphs • Human-in-the-Loop
  • •Architected a production multi-agent pipeline using LangGraph (without LangChain bloat) with stateful cyclic memory.
  • •Integrated real-time human-in-the-loop validation checkpoints for mission-critical automated workflows.

Voxket AI Platform & Voxket Connect

IVR Agents • WhatsApp API • Secured Webhooks
  • •Built agent tool-orchestration engines powering IVR conversational voice bots and automated WhatsApp pipelines.
  • •Deployed specialized autonomous workers including client meeting schedulers, real-time note takers, and secured HMAC webhooks.
System Performance: Optimized Django REST Framework APIs for speed and security: improved response times by 60% with query optimization, indexing, and Redis caching. Hardened auth and access control.
Core Trunk

Slashash Tech LLP

Full Stack / Node.js Developer

Dec 2022 – Jan 2025In-Office

“Financial RAG, Network Protocols & Secure SaaS Infrastructure”

FinGPT (Financial RAG Pipeline)

LangChain • OpenAI • Dynamic RAG
  • •Built a high-precision RAG pipeline for Indian tax advisory, featuring automated web scraping, data preprocessing, and strict prompt grounding.
  • •Eliminated hallucinations through continuous benchmarking and iterative system prompt engineering.

Onebuck DNS Resolver

Node.js • Networking Protocols • RFC Standards
  • •Implemented a custom network-oriented DNS resolver with granular user-level filtering rules.
  • •Built a low-latency centralized URL blacklist engine requiring deep exploration of DNS protocol packet structures.

Fiduciary Works (SaaS)

SaaS Architecture • docx.js • SQL Optimization
  • •Engineered core multi-tenant team and fee management modules with optimized relational queries.
  • •Automated programmatic PDF and DOCX financial compliance report generation via docx.js.

CashNGo Loan Engine

Python • Financial Algorithms • Fraud Detection
  • •Developed automated income assessment modules, compound interest accrual functions, and internal circular transaction detection.

Secure Share

Hybrid Cryptography • AWS S3 • Ephemeral Access
  • •Constructed hybrid-cryptography file encryption storing zero-knowledge chunks on AWS S3 with time-bound signed access tokens.
System Performance: Refactored backend data pipelines and database query plans, reducing complex reporting generation latency by 45% while enforcing zero-trust data access.
RootsRemote

SharmaStore

Frontend Developer

Apr 2022 – Jun 2022Remote

“E-Commerce Architecture & Responsive UI Engineering”

E-Commerce Platform Core

JavaScript • UI Architecture • Component Optimization
  • •Constructed responsive storefront layouts, product filtering logic, and high-conversion checkout flows.
  • •Diagnosed DOM re-render bottlenecks to deliver smooth navigation across mobile and desktop devices.
System Performance: Optimized client-side bundle sizes and critical asset loading paths, laying the foundation for high-performance web systems.
Technical Capabilities

Core Skills & Systems Expertise

Specialized in backend resilience, high-speed data access, and autonomous agentic AI systems.

Ordered Tech Stack Progression13 Core Competencies
01Python
02Django
03SQL Databases (MySQL / PostgreSQL)
04Next.js
05Node.js
06JavaScript
07MongoDB
08WebSockets
09RAG
010AI Agents
011LangChain
012LangGraph
013AWS Services (EC2, Lambda, S3)

AI & Autonomous Architectures

Building production-grade cognitive workflows, stateful multi-agent graphs, and grounded retrieval engines.

RAG (Retrieval-Augmented Generation)AI AgentsLangGraphLangChainLLM Fine-TuningModel Routing & Fallbacks

Backend & Systems Engineering

Engineering scalable microservices, low-latency APIs, and hardened authorization architectures.

PythonDjango & Django REST FrameworkFastAPINode.jsWebSockets & WebRTC SignalingSecured Webhooks (HMAC)

Databases & Storage

Data modeling, relational indexing, high-performance caching layers, and document stores.

PostgreSQLMySQLMongoDBRedis CachingQuery Optimization (60%+ latency drop)

Cloud & Web Runtime

Elastic infrastructure, compute provisioning, and modern full-stack integrations.

AWS EC2AWS LambdaAWS S3Next.jsJavaScript (ESNext)
Engineering Mindset

Problem Solving & Self-Directed Engineering

Beyond languages and frameworks, engineering excellence comes down to how quickly you diagnose unknown errors, optimize system architecture, and turn complex specifications into resilient software.

First-Principles Problem Solving

Deconstructing complex distributed issues into fundamental components to eliminate root bottlenecks rather than patching symptoms.

Smart & Hard Working Execution

Relentless work ethic paired with intelligent automation—designing systems that operate autonomously with minimal operational toil.

Self-Directed Deep Research

Demonstrated ability to master cutting-edge paradigms (e.g. DNS protocol internals, LangGraph cycles, model fine-tuning) autonomously.

Problem Solving and Engineering Mindset
Independent Systems & Engineering

Personal Projects

Deep-dive implementations highlighting autonomous agents, model fine-tuning, dynamic RAG, and distributed protocols.

Project 01 of 02AI Systems & Model Routing
Live DemoBackend GitHubFrontend GitHub

Custom AI Tool Builder & Indian Tax Chatbot

Autonomous Tool Generation Engine & Fine-Tuned Legal Intelligence

Generates dynamic schemas and input forms from plain English descriptions while isolating LLM usage purely to decision-making.

Qwen-3 Fine-TuningDynamic RAGLangChainTavily Search APIHugging FaceSecured Webhooks (HMAC)Model Fallback Routing

Key Architecture & Characteristics:

  • •LLM-for-Decisions Architecture: Decoupled reasoning from deterministic execution. All heavy work (HTTP API calls, transactional emails, webhooks) executes through general tools without wasteful token loops.
  • •Cost-Optimized Model Routing: Dynamic heuristic router switching between reasoning models and lite models, backed by automatic failover routing when primary endpoints fail.
  • •Fine-Tuned Qwen-3 Tax Chatbot: Fine-tuned Qwen-3 model trained on Indian Income Tax and GST legislation for individuals and business owners. Deployed on Hugging Face.
  • •Dynamic RAG with Real-Time Search: Augmented with Tavily and Google Search integrations for live grounding against up-to-date tax amendments and circulars.
  • •Secured Webhooks & Integrations: Endpoints protected with cryptographic HMAC signatures for external webhook dispatches and trigger events.
Complex Problem Solved: Solved the high cost and latency of traditional LLM apps by enforcing deterministic tool execution, caching query embeddings, and implementing automated fallback channels.
Custom AI Tool Builder & Indian Tax Chatbot
Tested for Zero-Downtime Resilience
Project 02 of 02Distributed Systems & Vision ML
Live DemoSource on GitHub

Real-Time AI Interview Proctoring System

Low-Latency WebRTC Mesh & Distributed Computer Vision Engine

Decentralized peer-to-peer proctoring pipeline with sub-100ms signaling and client-side distraction detection.

WebRTC MeshMediaPipe VisionSocket.IOReactSQLiteLow Latency SignalingTamper-Evident Logs

Key Architecture & Characteristics:

  • •WebRTC Mesh Topology: Optimized P2P request-response signaling handshakes to minimize central relay bandwidth and maintain ultra-low latency audio/video feeds.
  • •MediaPipe Vision Distraction Detection: Embedded computer vision models tracking head yaw/pitch, gaze vectors, and background multi-person detection directly in the browser.
  • •Timestamped Event Recording: Synchronized telemetry and violation alerts streamed via Socket.IO directly to an indexed SQLite audit store.
  • •Resilient Connection Recovery: Automatic ICE restart and signaling reconnect loops that prevent session disconnects over unstable network connections.
Complex Problem Solved: Eliminated expensive central video transcoding servers by delegating vision inference to client edge devices and orchestrating direct peer-to-peer WebRTC mesh connectivity.
Real-Time AI Interview Proctoring System
Tested for Zero-Downtime Resilience
Academic Background

Education & Academic Foundation

Continuous technical refinement combining formal academic degrees with relentless real-world production engineering.

In Progress

Master of Computer Applications (MCA)

Mangalayatan University

Currently Pursuing (Distance Learning Mode)

Advanced Computational Theory, Distributed Systems & Artificial Intelligence

Formal DegreeAccredited
Completed

Bachelor of Computer Applications (BCA)

Completed

Graduated

Computer Science Foundations, Relational Databases, Operating Systems & Object-Oriented Design

Formal DegreeAccredited
Initiate Collaboration

Let's Build Resilient Systems Together

Whether you need autonomous AI agents with LangGraph, a hardened Django REST backend, or sub-second API performance tuning, I'm ready to architect solutions.

Direct Channels

Reach Out Directly

Active response time within a few hours. Available for contract consulting, freelance architecture, and full-time engineering roles.

WhatsApp / Phone

+91 9468579471

GitHub Profile

github.com/naraina20

Location

Wadgaon Sheri, Pune, India

Open to Worldwide Remote Roles

Start WhatsApp Conversation
We Are Open - Contact Narayan Dewasi

Always open for intriguing backend & AI system challenges.