Tanisha Bhayani

• Built agentic pipelines using LangChain with tool integrations, multi-step orchestration, and LLM-powered prompting agents for similarity search and ranking across podcast Q&A, resume-job matching, research indexing, and legal meeting minutes generation.
• Developed a React front end with an analytics dashboard for real-time financial insights, complementing the Django and PostgreSQL back end that handles multi-currency invoicing, payment tracking, and transaction-safe reconciliation.
• Implemented n8n automations integrated with Gmail and Slack for fault detection, security alerts, and payment notifications on a crypto AI and quantum trading platform; built data visualizations showing the crypto market state and managed DevOps.
• Created the backend system for training GPT models on user’s custom dataset, while preserving privacy of the user-data, more like today’s private custom GPT on chatGPT, but user can fine-tune those GPT based on their dataset, deployed on Kubernetes for scalability, devised pricing plans for the startup considering the cost of the deployment per user

Computer Vision Quality Inspection (automotive manufacturing)

• Architected an AI-powered visual defect detection system for automotive manufacturing, defining AI/ML architecture, evaluating industrial cameras, and developing real-time object detection pipelines integrated with plant systems

AI Consulting (NLP, CV & RL) (Consulting)

• Delivered end-to-end AI solutions for Fortune 20 companies, startups, and enterprise clients across NLP, computer vision, and reinforcement learning
• Built multilingual BERT/GPT models, a financial NER model (99% F1) and property valuation models achieving <3% MAPE while mentoring client engineering teams

Circana (market research)

• Built a multimodal RAG platform over proprietary datasets with streaming responses and evaluated multiple LLMs to optimize response quality.

Schenck Process (industrial process technology)

• Built a React application for image annotation and trained lightweight object detection models optimized for mobile deployment

AI Trading Platform (Fintech)

• Developed a Bloomberg-style trading interface with LLM-powered market analysis, dynamic charting, automated operational workflows, and led architecture validation

Sign Language Recognition (AI startup)

• Developed continuous and isolated Arabic Sign Language recognition pipelines, including transformer-based models, video segmentation, and end-to-end inference APIs

AI Meeting Minutes

• Built a multi-tenant AI meeting platform from scratch, including transcription, RAG indexing, and automated minutes generation, achieving an 80% BERTScore against human-written summaries

E-Learning Platform

• Architected an AI-powered learning platform featuring course recommendations, SAT-style question generation, ingredient segmentation, calorie estimation, and supporting backend APIs

• Teaching MSC-IT Students Machine Learning course. Teaching concepts like Clustering, Recommendation Systems, semi-supervised Learning, and ensemble methods.