Driving AI & Automation @ Dealshare · ex-Reliance Retail · IRMA
I design and ship AI-powered products end to end — from data pipelines and LLM workflows to clean web apps and 100+ production automations. Idea → built → live.
Get in touch

A full recruiter platform: positions organised into hiring "workbooks", candidates auto-screened by a bilingual English/Hindi AI voice line (on Plivo, ~₹3.60/min), WhatsApp re-engagement after missed calls, resume scoring, and a live candidate board with call/interview status — Interviewed, DNC, scheduled, and more.

A topic in, a finished YouTube episode out: scripted narrative, AI voiceover, character-locked AI visuals, forced-aligned chapters and a full publish kit (titles, thumbnails, metadata) — assembled automatically. It runs the live channel The Long Story Hindi.

A staff console where an agent launches an AI voice call that conducts the complaint conversation in Hindi — capturing, classifying and routing the issue end to end, with the AI handling the back-and-forth.

Upload a CV, get a sharper, recruiter-ready version in ~15 seconds — rewritten copy, a typeset PDF, and a locked professional layout. First scan free, no account needed.

An interactive map tool for retail catchment analysis — WorldPop population heatmaps layered with OpenStreetMap points of interest, to size and compare a store's trade area block by block.
LangChain agents that answer business questions over live Postgres/BigQuery — including an automated margin-dip analysis that's triggered by webhook, reasons over the data, and posts findings back to the team.
Scheduled Python scrapers of internal warehouse systems that catch out-of-stock, GRN and dump anomalies the moment they happen and push formatted alerts & rendered PNG report-cards to Discord — later migrated to WhatsApp via the Evolution API.
A dedicated FastAPI micro-service that scrapes Blinkit through bot-protection (cloudscraper + rotating residential proxies), plus Flipkart Minutes and DMart private-label benchmarking — normalised to CSV and pushed to WhatsApp/Sheets as daily price & stock intel (built for XYXX).
A WhatsApp AI assistant that handles text, voice notes, images and PDFs — transcribes and reasons with OpenAI/Gemini, replies in synthesized voice via ElevenLabs, and keeps per-user memory in Postgres.
Pulls live NSE stock and index data on a schedule into a Google-Sheets dashboard — a lightweight personal markets tracker driven by Python + pandas.

A daily decision system for a ~20-store grocery chain: BigQuery→Postgres pipeline, demand forecasting with auto-generated indents, like-to-like weekly analysis, and an LLM "owner playbook" that turns the numbers into the day's priority actions.

A short, warm chat about someone you love becomes original lyrics, sung into a real track in minutes — interview by one model, lyrics by another, vocals generated, then a pay-gated MP3 + lyric PDF.

Drop a product sheet (Excel/CSV) and a folder of images; Canvas matches them, resolves an intent-based layout (not fixed pixels), and exports on-brand marketing creatives across sizes — at scale.

A personal planner you talk to: tell it "I just woke up" or "I'm out today, no laptop" and it reshuffles your day around your energy and context. Installs to the iPhone home screen.
Working with earth-observation data for population, land-use and footfall analysis — Google Earth Engine and ISRO's Bhuvan, VIIRS night-time lights as an activity proxy, and Landsat / Sentinel multispectral imagery — turning raw satellite data into retail site & catchment insight.

A block-by-block visualization of Mumbai's 2024 rental market — where demand and price concentrate across the city, rendered in QGIS from open data.

Turning decades of Indian-cinema data into a visual story — a data-storytelling piece built with Python and Tableau.

A data-driven look at the quick-commerce battle in Mumbai — comparing Blinkit and Zepto on the metrics that actually move the market.

How a stack of AI + n8n automations can do the work of a full-time role for a few hundred rupees a month — the thinking behind much of the automation work above.
A product idea, a messy data problem, or a workflow that should be automated? Let's hop on a call and figure it out together.