Full archive

Work that
shaped
how
I think.

A working library of applied AI, automation, retail learning, MBA notes, and market observations. Each item is framed by the problem, the decision it supported, and the lesson I still carry from it.

Items

Applied / built

Living system

Showing 16 of 16 archive itemsFramed by problem, business signal, and learning.

Abstract system map for the living AI portfolio workflow

Living system map

AI-Assisted Product BuildingBuilding

Living AI Portfolio System

A reviewed, AI-assisted portfolio system that keeps proof, MBA notes, analytics, and public writing in one living loop.

Changed

Created a maintainable portfolio system with project archiving, journal foundations, analytics reporting, LinkedIn adaptation, and approval-first automation.

Took away

AI is useful when the human direction is clear. Taste, review, and positioning still decide whether anything feels credible.

Product direction · Content strategy · Next.js

Image-led search

AI + DATA SCIENCEShipped

Image-Led Product Discovery

Image-led product discovery flow for an agri-commerce context, helping shoppers move from visual intent to product language.

Changed

Increased orders by 28%.

Took away

Applied AI becomes valuable when it removes effort at a specific moment in the purchase journey.

OpenAI · FastAPI · Image processing

Operational alert timeline for order drop detection

Signal timeline

Automation & OperationsOperational

Order Drop Early Warning

Hourly anomaly layer for BigHaat order movement, designed to surface unusual shifts earlier in the day.

Changed

Created a faster signal layer for operational response instead of waiting for end-of-day diagnosis.

Took away

Decision support matters most when it connects signals to urgency and clear follow-up decisions.

Python · Anomaly detection · Dashboards

Retrieval flow

AI + DATA SCIENCEBuilt

Faster Product Lookup with RAG

RAG-style lookup for product and internal PDFs so repeated document questions did not keep becoming manual work.

Changed

Reduced manual search effort and improved speed of internal information access.

Took away

Retrieval work is useful when it makes internal knowledge easier to reach at the moment of need and still keeps trust visible.

LLMs · RAG · Python

Extraction flow

Automation & OperationsBuilt

Invoice Data Extraction Automation

OCR and parsing workflow that turned repeated invoice and PDF extraction into a structured review task.

Changed

Reduced repeated manual data-entry and validation effort.

Took away

Practical automation becomes valuable when it removes repeated effort without hiding the places where human review still matters.

EasyOCR · pdfbuilder · Python regex

Pricing signals

Business IntelligenceBuilt

Competitor Pricing Intelligence

Crawled 9 competitor sites, matched products fuzzily, and surfaced price movement as a business signal.

Changed

Identified competitor price changes mirroring ours by 65%.

Took away

Pricing is not only a technical tracking problem. It is a market signal with timing, imitation, and competitive pressure inside it.

Selenium · Scrapy · PostgreSQL

Status flow

Automation & OperationsBuilt

Logistics Visibility Automation

Logistics scraping workflow that turned fragmented partner status updates into clearer customer communication triggers.

Changed

Improved status visibility and supported more timely customer communication.

Took away

Customer communication improves when operational visibility improves first.

Web scraping · Automation workflows

Content cadence

Automation & OperationsShipped

Content Operations Automation

Automated content collection workflow that made publishing more consistent and useful for a farming audience.

Changed

Increased engagement by 144%.

Took away

Content operations need both relevance and cadence; one without the other fades quickly.

Web scraping · Automation

Advisory loop

AI + DATA SCIENCEImproved

Crop Advisory Model Improvement

LLM and image-processing work aimed at making a crop advisory workflow more useful in high-context farm decisions.

Changed

Improved the practical usefulness of an advisory workflow; exact public metrics are not shown here.

Took away

AI in agriculture needs reliability, context, and practical usefulness before it deserves trust.

LLMs · Image processing

Prediction signal

AI + DATA SCIENCEBuilt

Cancellation Risk Prediction

Prediction model for identifying likely order cancellations earlier, so follow-up could happen with better timing.

Changed

Reduced manual confirmation delay by 1-2 days.

Took away

Predictive work matters when it improves timing, reduces uncertainty, and gives teams a clearer next action.

Machine learning · Python

Reliability loop

Automation & OperationsImproved

Production Workflow Optimization

Production Python improvements that reduced runtime, made failure modes more visible, and strengthened operational reliability.

Changed

Reduced production runtime by 60%.

Took away

Performance is not just engineering polish; it changes whether a workflow feels dependable to the people using it.

Python · Database optimization · Alerts

Screenshot slot

B.Tech ProjectsAcademic build

Collaborative Diagramming Tool

Real-time collaborative diagram platform with document sharing.

Changed

Created early full-stack experience across collaboration, auth, and real-time architecture.

Took away

Early experience in real-time collaboration, authentication, and full-stack architecture.

Node.js · Express.js · Socket.IO

Screenshot slot

B.Tech ProjectsAcademic build

StockSense Forecasting Interface

Stock price prediction app with forecasting and interactive visualizations.

Changed

Created a usable interface for exploring time-series forecasts and market data.

Took away

Forecasting becomes useful only when the output is interpretable.

Python · Streamlit · Prophet

Screenshot of the Leather Defect Detection upload interface

Live app capture

B.Tech ProjectsAcademic build

Leather Defect Detection

Deep learning ensemble model for leather defect detection with Flask integration.

Changed

Created experience in model comparison, backend integration, and API-based deployment.

Took away

Model accuracy depends heavily on environment, data quality, and deployment context.

Python · CNN · VGG

Observation notes

Retail LearningLearning

CM Silks Retail Learning

A family saree retail lens on trust, aspiration, local positioning, pricing, customer education, and modern brand memory.

Changed

Built practical exposure to customer education, local brand memory, and trust-led retail communication.

Took away

Retail keeps reminding me that many purchase decisions are emotional, social, and practical at the same time.

Content planning · Customer observation · Retail positioning

Case space

MBA Case NotesBuilding

MBA Case Notes & Market Signals

A growing space for product, marketing, strategy, consumer behavior, and market analysis work from the IIM Sirmaur chapter.

Changed

Future-ready structure for adding stronger MBA work as coursework, cases, live projects, and market notes develop.

Took away

This section should grow from real coursework, cases, observations, and projects rather than filler.

Product thinking · Marketing · Strategy

Work that shaped how I think.

Living AI Portfolio System

Image-Led Product Discovery

Order Drop Early Warning

Faster Product Lookup with RAG

Invoice Data Extraction Automation

Competitor Pricing Intelligence

Logistics Visibility Automation

Content Operations Automation

Crop Advisory Model Improvement

Cancellation Risk Prediction

Production Workflow Optimization

Collaborative Diagramming Tool

StockSense Forecasting Interface

Leather Defect Detection

CM Silks Retail Learning

MBA Case Notes & Market Signals

Work that
shaped
how
I think.