[SYSTEM SPECIFICATION / OPERATOR PROFILE]

FULL-STACK, MOBILE & ON-DEVICE AI ENGINEER.

Mechanical Engineering student at Bangalore Institute of Technology (BIT) building and shipping production systems — real-time audio/video WebRTC gateways, on-device AI LLM pipelines, and high-concurrency mobile apps.

[OSCILLOSCOPE & SPECTRUM ANALYZER]01 FULLSTACK

CODE:TS / NEXT.JS / NODE / POSTGRES

SPECS:REUSABLE SCAFFOLDS · FAST DELIVERIES

FREQ: 44.0 kHzGAIN: +12 dBMOD: 65 %PULSE: #0

[ANALOG CONTROLS]

STANDBY

LINKEDIN GITHUB TWITTER EMAIL

[CATALOG / SHIPPED PRODUCTS]

Selected Work

VIEW ALL RECORDS →

[01]MARKETPLACE / INFRASTRUCTURE2024

Cleriko

Structured task marketplace replacing WhatsApp legal coordination. Razorpay escrow payments, JWT authentication, Socket.IO real-time bidding, and a custom Android SMS relay service in Kotlin ($0 per OTP cost).

React Native (Expo)TypeScriptNode.jsPrisma+4

VIEW SPEC →

[02]REALTIME / TELEPHONY2024

Fancall

Celebrity calling platform supporting live paid voice and video sessions. Built a Go WebRTC gateway with Pion bridging WebRTC audio to SIP endpoints via Vobiz. Solved iOS AVAudioSession CallKit audio drops and fixed ghost calls.

Swift (iOS)Kotlin (Android)GoPion WebRTC+4

VIEW SPEC →

[03]ON-DEVICE AI / MOBILE2025

Haven AI

Voice-first productivity app with 100% on-device AI. Quantized Qwen 2.5 0.5B and Whisper running locally via llama.rn, automated task extraction, natural language date parsing, and local semantic vector search on 4GB RAM devices.

React Native (Expo)TypeScriptllama.rn (Qwen 2.5 0.5B)whisper.rn+1

VIEW SPEC →

[04]OPEN SOURCE / AI PROTOCOL2025

Blender MCP Server

Adopted into Cursor IDE's official MCP catalog. Author of an open-source Model Context Protocol server exposing Blender 3D as a tool-callable API surface for LLM agents to generate scenes, edit meshes, and render procedurally.

PythonModel Context Protocol (MCP)Blender Python APICursor IDE

VIEW SPEC →

[05]HARDWARE / INFERENCE2024

Local AI News Summarizer

Repurposed a 7-year-old Android phone in Termux as a self-hosted AI inference server running Qwen 1.5B via llama.cpp. Automated RSS news pipeline surfaced in a TikTok-style vertical feed reaching 1,000+ daily active users.

Pythonllama.cppQwen 1.5BTermux+2

VIEW SPEC →

[06]ENGINEERING / FREELANCE2024–Present

Software Engineer (Freelance)

Shipped 5+ production applications across mobile, web, blockchain, and AI. Built reusable full-stack TypeScript scaffolds and authored Solidity smart contracts for trustless escrow payment flows across client engagements.

TypeScriptReact NativeNext.jsNode.js+3

VIEW SPEC →

[07]FINTECH / QUANT2025

Nifty Strats

NIFTY 50 options analysis toolkit featuring 30+ trading strategy signal generators, dark-themed option chain visualizer, Black-Scholes P&L payoff calculator, and an LSTM direction classifier in PyTorch.

PythonPandasPyTorchPlotly+2

VIEW SPEC →

[08]DATA RESEARCH2025

Fresher Hiring in India

Empirical research analyzing 5,000 fresher job applications across 68 companies and 14 sectors. Uncovered salary distributions, hiring funnels, skill premiums, college tier myths, and what actually drives job offers.

PythonPandasMatplotlibSeaborn+1

VIEW SPEC →

[SYSTEM ARCHITECTURE / CAPABILITIES]

Technical Skills Matrix

[LANGUAGES]

TypeScriptJavaScriptPythonGoSwiftKotlinSQLC++

[BACKEND & FRAMEWORKS]

Node.jsExpress.jsReact Native (Expo)Next.jsReactPrisma ORMZod

[MOBILE & REAL-TIME]

iOS (Swift, CallKit)Android (Kotlin)WebRTC (Pion)SIP / VoIPSocket.IO

[AI & ON-DEVICE INFERENCE]

llama.cppllama.rnwhisper.rnQwenMCP (Model Context Protocol)Semantic Search

[DATABASES & INFRA]

PostgreSQLRedisSQLiteMongoDBDockerGitHub ActionsVercelLinux

[INITIATE TRANSMISSION]

Let’s Build Something Exceptional.

Available for freelance engineering, co-founding, or full-stack roles. If you build fast and care about design & technical craft, let’s connect.

contact@kushalneversleeps.com →