hi, i am hackice20!

Full-stack engineer

i am a 22.36290873 yr old engineer, building in Web Dev, Agentic AI & ML.

i like tight loops from idea to production: apis, ui, and ai features that stay fast, secure, and boring to operate.

work

  • Kelp · Internship

    Full Stack Engineer

    Mumbai, Maharashtra, India · On-site

    Dec 2025 - May 2026 · 5 mos
    • Architected deals tool module integrating Elasticsearch queries across 4 data sources with intelligent parameter mapping for 2 enterprise clients.
    • Implemented signed URL mechanism with Redis caching (15-minute TTL) to secure file downloads, eliminating filepath exposure vulnerabilities and preventing unauthorized file enumeration attacks.
    • Refactored redundant JWT authentication across main and external investor portal into a unified auth service with dynamic database routing, eliminating code debt.
  • Magnacamz · Internship

    Full Stack Engineer

    Pimpri Chinchwad, Maharashtra, India · On-site

    Dec 2024 - Jan 2025 · 2 mos
    • Architected and maintained 3 full-stack web applications, including 2 production-grade product websites and 1 interactive project demo.
    • Integrated rate limiting, caching, and optimized API routes to boost backend performance by 40% while ensuring scalable and secure data handling across user sessions.
    • Crafted responsive and SEO-optimized UIs with Tailwind CSS and Vite, reducing page load time by 35% and improving Lighthouse performance scores (90+).

blogs

stuff i built

QueryNox

QueryNox

  • Multi-model chat (OpenAI, Claude, Gemini, Llama) with web search augmentation and document RAG.
  • Subscriptions and billing through Polar with plan limits and usage gating.
  • Image generation and user uploads stored on Cloudflare R2; PDF and multi-file ingestion for context.
  • Observability across API requests, errors, and latency; session-scoped chat history and custom system prompts.
Multi-LLMRAGPolarCloudflare R2Observability
  • Chat-to-app builder: generate and preview full-stack apps from prompts in under 60 seconds.
  • Multi-provider LLM streaming with SSE, plan/build modes, and mode-policy enforcement; 11-entity Postgres schema with advisory locks.
  • In-browser WebContainer sandbox with dependency snapshots (IndexedDB + Supabase Storage) so installs are not repeated every session.
  • Terminal telemetry and auto-recovery with runtime event logs and audit trails.
WebContainerSupabaseSSEMulti-LLMAgentic

Shipping more soon...

git graph

GitHub contribution graph for hackice20
“I would rather be ashes than dust.”

contact

© 2026 yashworks.com