I'm an AI & Data Science engineer and independent developer in Bengaluru, building under MYK Productions. B.Tech in Artificial Intelligence & Data Science, class of 2026.
I take products the whole way — computer vision and speech models, agentic LLM apps, and real-time multi-tenant SaaS that paying clients run in production.
- Building — BlockSense, a real-time operations platform for gated communities (web + installable PWA)
- Going deeper on — generative AI, LLM tooling, and getting models to survive production
- Freelancing — shipped and secured client systems: Supabase RLS, edge functions, CSP/HSTS hardening
- Open to — ML engineering, data science and software roles, plus applied-research collaborations
- Beyond code — led 300+ student ambassadors on national campaigns for Spotify, Netflix and Pepsi
Multi-tenant SaaS for gated communities. A 30+ table schema across three role-based portals, live water/power/gas/sewage monitoring, and a prediction engine that forecasts depletion from 7-day rolling averages — firing WhatsApp and email alerts before a tank runs dry.
Live demo · Mobile PWA — ~100 ms real-time sync, offline-capable | AI personal-finance dashboard. Parses Indian bank SMS into a categorised ledger in real time, detects recurring subscriptions, forecasts your balance — and runs an agentic chat that executes actions: creating goals, setting budgets, firing overspend alerts.
|
Classifies 7 emotions from voice by fusing a neural net with pitch analysis, NLP sentiment on transcribed speech, and voice heuristics in a weighted ensemble. Ships explainable output — which of pitch, energy or word choice drove each call — across a Streamlit app, a Python API and a CLI.
| Real-time vacant-bay detection from parking-lot video. Grayscale → Gaussian blur → adaptive thresholding, then pixel-density analysis per bay, with colour-coded boxes, a custom space-picker tool and a live spot counter. No GPU required.
|
A production QR-to-WhatsApp lead funnel for a fashion retail brand: Supabase Postgres with row-level security, a Deno edge-function write path with rate limiting, dedup and bot protection, an admin dashboard with CSV export, and per-campaign QR attribution. Hardened with CSP/HSTS, XSS and CSV-injection protection.
| A code-driven video pipeline — renders a multi-scene product explainer entirely from TypeScript components via Remotion, with a reusable theme system and an automated render script. Front-end built on TanStack Start.
Private — happy to walk through it on request |
AI / ML
Backend & Data
Frontend
Ship & Run
