AI Portfolio
Building with AI
before it was obvious.
Not experiments. Not demos. Production AI products — shipped, live, and solving real user problems.
All products live in production
Real users — not synthetic evals
Measurable outcomes — not vanity metrics
AI approaches used
LLM-powered intent detection
CS Automation
Agentic AI & multi-step workflows
sahaAI Bot
Image recognition & attribute extraction
Catalog Creation
Vector embeddings & semantic search
Semantic Search
Shipped products
The work, in detail
Live in production
Customer Support
CS Automation
Resolving customer queries without human intervention
~80%
Queries resolved autonomouslyAI detects intent, routes, responds and resolves — integrated with refunds, cancellations & ticketing
Chat & email support fully automated end-to-end
Integrated with order workflows & ticketing systems
Intent DetectionLLMWorkflow AutomationNLP
Live in production
Conversational Commerce
Conversational Transaction Bot
sahaAI
Discover, order and checkout — through conversation
0→cart
Full transaction via text or voiceAgentic AI handles search, discovery, add-to-cart and checkout — no traditional UI required
Supports text & voice for food discovery & ordering
Multi-step agentic workflows with full transaction support
Agentic AIVoiceConversational UXMultilingual
Live in production
Seller Tools
AI-Assisted Catalog Creation
From a single photo to a marketplace-ready listing
1photo
Auto-generates full product listingAI extracts attributes, generates title, description & category — eliminating manual catalog effort
Standalone seller app — photo in, listing out
Drastically reduces seller onboarding friction
Image RecognitionGen AIAttribute Extraction
Live in production
Search & Discovery
Semantic Search
Understanding what users mean, not just what they type
~98%
Zero-result queries eliminatedReplaced keyword matching with vector embeddings — users find products even with natural, conversational queries
Intent-aware search across a sparse catalog
Dramatically improved search coverage & relevance
Vector EmbeddingsNLPIntent ModelingSemantic Ranking
How I think about AI in product
The philosophy
behind the work
behind the work
01
Problem first, AI second
I start with a real user problem — AI is the tool, not the brief. If a simpler solution works, I use that.
02
Production beats prototypes
A live product with 100 users teaches you more than a polished demo shown to 1000 stakeholders.
03
Measure the right thing
AI features need outcome metrics, not model metrics. I care about the 80% resolution rate — not the F1 score.
On the horizon
Let’s build
Want to build
something together?
Open to consulting, advisory roles, and senior product leadership opportunities.