14-year-old systems developer based in Pune, India. Designing private multi-agent pipelines, deploying self-hosted models on Linux VPS, and architecting sub-30ms edge applications with ₹0 recurring cloud bills.
Production-grade software systems, audited calculation engines, and autonomous agent loops engineered under strict performance and privacy constraints.
High-speed client-side calculation engine for Maharashtra motor vehicle taxes. Accurately computes complex state slabs across vehicle classes, fuel types, and import categories with zero server dependencies.
High-throughput local AI proxy routing inference requests dynamically between local self-hosted models (Ollama/vLLM) and upstream frontier models. Features intelligent fallback, token streaming, and request sanitization.
Autonomous background execution loops powered by Hermes Agent and n8n workflows. Connects self-hosted coding models (Qwen 2.5 Coder 14B) with Telegram bots, automated code review triggers, and webhook event dispatchers.
High-performance edge deployment architecture running on Cloudflare Pages and Workers. Handcrafted vanilla HTML5/CSS/JS without heavy framework baggage, optimized for sub-30ms global response and complete AEO/SEO indexing.
Client deliverables, platform optimizations, and automotive AI diagnostics engineered for partner companies.
Concrete technical deliverables, challenge breakdowns, and verified production outcomes.
Architected an automated multi-agent triage routing system connecting vehicle Diagnostic Trouble Codes (DTC) with local LLMs and edge fallback pipelines, translating raw telemetry into actionable technician guidance.
Sub-120ms diagnostic query response; eliminated runaway third-party API costs with self-hosted local model routing.
Engineered the web platform and high-speed media delivery pipeline for India's premier motocross and dirt biking community. Optimized asset compression, mobile responsiveness, and event registration flows.
Under 100ms mobile page load times across Indian 4G/5G networks; 100% uptime during high-traffic national race events.
Dalvi Consultancy
Built autonomous backend webhook pipelines, automated client intake flows, and internal business logic in n8n and Linux VPS daemons, eliminating manual data entry bottlenecks.
Saved dozens of hours of manual administrative operations monthly with self-healing scheduled automation scripts.
Architected high-speed static edge sites utilizing modern web standards without heavy JavaScript framework runtime bloat, targeting sub-30ms first-contentful paint globally.
Perfect 100/100 Google Lighthouse scores across all audits with ₹0 monthly hosting bills via Cloudflare edge caching.
I'm 14, based in Pune, India. I started programming at age 11 running a 30-player PaperMC server, which pushed me straight into managing Linux VPS environments, Docker containers, and database backends out of sheer necessity.
Rather than burning cash on proprietary cloud APIs, I engineer local-first: deploying self-hosted models (Ollama, vLLM) on hardened Linux VMs, paired with Cloudflare Workers for sub-30ms edge delivery with ₹0 recurring cloud bills.
Offline, I balance terminal time by riding dirt bikes (KLX 230r), trekking 10km mountain trails across the Sahyadris with GGIM, and studying natural habitat architecture. Physical constraints teach systems thinking better than any tutorial.
Self-hosted model clusters, Linux VPS daemons, and autonomous workflow pipelines.
Multi-modal reasoning with Claude 3.5 Sonnet, GPT-4o, and Gemini 1.5 Pro for deep architectural planning, system verification, and code synthesis.
Private local inference using Ollama and vLLM (Qwen 2.5 Coder 14B, DeepSeek, Llama 3) running on Linux VPS with zero data leakage.
Hardened Debian/Ubuntu cloud VMs managing 24/7 background daemons, Docker containers, reverse proxies, and headless agents.
Autonomous multi-step webhook automations, scheduled background tasks, and function-calling integrations connecting APIs and LLM backends.
Autonomous CLI execution loops utilizing ReAct frameworks, tool-calling pipelines, persistent memory, and self-correcting error handling.
Automated CI/CD pipelines, version control, automated linting, test suites, and production edge deployments via Cloudflare Pages.
Test real command outputs or click chips below to inspect system telemetry, builds, and architecture decisions.
Open for technical collaboration, autonomous AI agent architecture, and high-performance edge contracts.
Direct inbox for technical contracts, agent workflows, and architecture consultations. Typically responding within 24 hours.
Your message has been formatted. Click below to launch your email client or copy directly to inbox.