Mohit Mathur profile avatar Mohit Mathur
Hire Me

Selected work

Flagship case studies

Filter:

AI-Powered Sales & Inventory Analytics Chatbot

Jan 2025–Aug 2025

Business Context
Business owners needed a faster way to understand sales, inventory, stock movement, and business trends.
Role
Led a 4-person cross-functional team and the platform delivery.
Team
4 people across AI engineering, backend, and QA.
Scope
Conversational analytics for sales, inventory, stock movement, trends, and recommendations.
Delivery Model
Built and shipped the analytics platform.
Challenge
Reduce manual reporting, stock-out incidents, and the time needed to make decisions.
Actions
Delivered natural-language interactions and actionable recommendations over business data.
Outcomes
70% less manual reporting; 35% fewer stock-out incidents; 60% faster decision-making.
Technology
Python, FastAPI, OpenAI API, LangChain, MySQL, MongoDB, Vue.js, ClickUp.
Evidence
Request project evidence
Case Study

Inventory Listing Solution

Oct 2022–Present

Business Context
The business needed centralized inventory listings across its channels.
Role
Directed the 11-person team through requirements, delivery, risk planning, stakeholder communications, and deployment.
Team
11 people.
Scope
Centralized inventory listing platform.
Delivery Model
Requirements through deployment, with risk planning and stakeholder communications.
Challenge
Improve inventory accuracy and reduce manual inventory errors.
Actions
Directed delivery of a centralized inventory listing platform.
Outcomes
40% higher inventory accuracy; 35% fewer manual inventory errors; 10% under budget and 2 weeks ahead of schedule; 95% client satisfaction.
Technology
CFML, Coldbox, Vue.js, AWS, Flutter, WordPress, ClickUp.
Evidence
Request project evidence
Case Study

E-commerce Multi-Vendor Portal

Jun 2022–Nov 2022

Business Context
The business needed centralized inventory management and vendor-managed product listings.
Role
Led delivery with a 5-person team.
Team
5 people.
Scope
Multi-vendor portal for centralized inventory and vendor-managed product listings.
Delivery Model
Agile delivery cycles, with scope, roadmap, and risk mitigation.
Challenge
Improve inventory data accuracy and reduce manual product updates.
Actions
Led delivery of a portal that centralizes inventory and supports vendor-managed listings.
Outcomes
30% better inventory data accuracy; 50% fewer manual inventory updates.
Technology
Laravel, AngularJS, MySQL, ClickUp.
Evidence
Request project evidence
Case Study

Healthcare Staff Management Service Provider

Jan 2024–Jun 2024

Business Context
A staffing provider needed a system to support recruitment, onboarding, and staff operations while lowering reliance on costly third-party tools.
Role
Managed the full project lifecycle.
Team
Team size not specified in the supplied details.
Scope
Planning, requirements, execution, and post-launch support for a staffing system.
Delivery Model
Full lifecycle delivery, from planning through post-launch support.
Challenge
Improve recruitment and onboarding speed, operating cost, client satisfaction, and staff utilization.
Actions
Managed delivery of the staffing management system.
Outcomes
30% faster recruitment and onboarding; 25% lower operating cost; 15% ahead of schedule; 20% higher client satisfaction; 40% better staff utilization.
Technology
Jira, CFML, Coldbox, Vue.js, MySQL.
Evidence
Request project evidence
Case Study

Proactive Risk & Dependency Framework

Dates not specified

Business Context
Improve delivery predictability by reducing delays related to project risks and dependencies.
Role
Role details not specified in the supplied project information.
Team
Team size not specified.
Scope
Proactive risk and dependency framework.
Delivery Model
Not specified.
Challenge
Reduce delivery delays.
Actions
Applied a proactive risk and dependency framework.
Outcomes
20% reduction in delivery delays.
Technology
Not specified.
Evidence
Request project evidence
Case Study

AI/ML Lab · Public implementations & coursework

Learning by building

A hands-on record of implementations, experiments, and coursework. These projects are educational and experimental work, not production AI deployments.

GPUAI & MLResearch

CUDA implementation · coursework

CUDA Matrix Exponentiation

How much can exponentiation by squaring and shared-memory tiling speed up large matrix powers on a GPU?

  • 11.7×–28.3× GPU speedup across the three matrix sizes.
  • Maximum reported CPU/GPU output error: 0.00021.
View RepositoryBenchmark files
NLPAI & MLResearch

NLP coursework · fine-tuning experiment

Sanskrit → English NMT

Fine-tune IndicTrans2 for low-resource Sanskrit-to-English translation and inspect translation quality.

  • Validation evaluation uses BLEU and BERTScore; numeric scores are not published in the repository summary.
  • Produces held-out test predictions and lexical-overlap error analysis.
View RepositoryOpen Notebook
AI & MLResearch

From-scratch implementation · coursework

Decision Trees from Scratch

Implement entropy and Gini splits in NumPy, then examine pre-pruning and decision boundaries on Iris.

  • 96.7% test accuracy at depth 3 on the Iris dataset.
  • Includes Gini/entropy comparisons and a decision-boundary visualization.
View Repository
AI & MLResearch

From-scratch implementation · coursework

Random Forest from Scratch

Build an ensemble from custom decision trees with bootstrap sampling and random feature selection.

  • 79.8% test accuracy on the binary Wine Quality task with 50 trees.
  • Compares against a custom decision tree and reports feature importance.
View Repository
AI & MLResearch

Validation implementation · coursework

K-Fold Cross-Validation

Implement K-fold validation around a K-nearest-neighbors classifier using the Iris dataset.

  • Custom fold logic evaluates KNN performance across held-out folds.
  • Runnable Python script fetches Iris data and executes the evaluation workflow.
View Repository
AI & MLResearch

From-scratch implementation · coursework

Linear Regression from Scratch

Use NumPy gradient descent, optional L2 regularization, and held-out evaluation on California Housing.

  • Test R²: 0.5672 on California Housing.
  • Test MSE: 0.5672; includes a learning curve and actual-vs-predicted plot.
View Repository