About Me
Welcome to my portfolio! I'm Eesita Sen, a product-focused developer and AI enthusiast passionate about translating ideas into impactful, user-friendly solutions. My journey began five years ago, and I've since worked on everything from user experience design and testing to AI-driven chatbots and automated cloud infrastructures. I believe in blending user-centered design with rigorous software development practices to ensure that every product not only works seamlessly but also resonates with its audience. My goal? To build digital products that delight users at every interaction, and to do it with empathy, creativity, and precision.
In my spare time, I love exploring new culinary flavors, planning my next travel adventure, and indulging in sci-fi and lighthearted workplace & lifestyle dramedies.
Projects Overview
Below is a quick snapshot of some key projects I've worked on. Click "View Details" or the GitHub link for more info.
Jenoai.com - Marketing Analysis and Recommendation System
Analytics
FastAPI
AWS ECS
Built and deployed FastAPI microservice on AWS ECS to generate marketing reports from business name/location; used LLM caching to reduce load time and cost by 40%.
Key Metric: 85% classification accuracy with 90% R² predictions.
Jenoai.com - Marketing Analysis and Recommendation System
Duration: Jul 2025 - Aug 2025
Technologies: FastAPI, AWS ECS, LangGraph, XGBoost, Google Places API, Vector Search, LLM Caching
Overview:
Built and deployed FastAPI microservice on AWS ECS to generate marketing reports from business name/location; used LLM caching to reduce load time and cost by 40%.
Key Contributions:
- Workflow Orchestration: Used LangGraph workflow to generate the reports using Google Places data, ML models, web search, vector search, and LLM; parallelized nodes to keep response time under 60 seconds.
- ML Model Development: Trained XGBoost models on data from 5,000+ businesses (scraped, cleaned, labeled via KeyBERT) with 20+ features, achieved 85% classification accuracy and 90% R² on marketing and online presence predictions.
- Performance Optimization: Implemented LLM caching strategies to reduce API costs and improve response times by 40%.
Outcome:
Delivered a high-performance marketing analysis system that provides accurate business insights with fast response times and cost-effective operation.
AI-Based Portfolio Agent and Chatbot
AI
JavaScript
AWS
Developed an AI-powered chatbot leveraging OpenAI's LLM and RAG framework that analyzes URLs when pasted, extracting key information and providing responses to portfolio-related queries.
Key Metric: Reducing time-to-insight by 30%.
AI-Based Portfolio Agent and Chatbot
Duration: Jan 2025 – Feb 2025
Technologies: OpenAI LLM, RAG Framework, AWS, Node.js, JavaScript
Overview:
Developed an intelligent chatbot leveraging a Retrieval Augmented Generation (RAG) framework and OpenAI LLM to deliver real-time portfolio insights and responses, enhancing user engagement and information accuracy.
Key Contributions:
- AI Chatbot Development: Designed and implemented a RAG-based chatbot using OpenAI LLM and vector databases for dynamic, context-aware responses.
- Dynamic Data Extraction: Integrated web scraping tools (Cheerio, Axios) to automatically extract job descriptions and portfolio data, generating structured, AI-powered summaries.
- Robust Deployment & Security: Achieved zero-downtime updates on an EC2 instance by implementing automated deployment workflows with GitHub Actions, managing DNS with Route 53, and securing HTTPS via AWS Load Balancer + Certificate Manager.
Outcome:
Delivered a responsive and secure AI-driven portfolio agent that streamlined information retrieval, maintained high availability, and provided real-time, data-rich insights to users.
Wyckoff AI Assistant
AI
Python
Flask
Engineered a transformer-based AI chatbot using PyTorch and Flask, integrating attention mechanisms and tokenization.
Key Metric: 85% accuracy with sub-second latency.
Wyckoff AI Assistant
Duration: Jan 2025 – Mar 2025
Technologies: PyTorch, Flask, Python, REST API
Overview:
Engineered a transformer-based AI chatbot using PyTorch and Flask, integrating attention mechanisms, tokenization, and REST API architecture to process 1000+ Wyckoff trading concepts with 85% accuracy and sub-second latency.
Key Contributions:
- AI Model Development: Implemented intelligent error handling and fallback response logic in Python, reducing failures by 95% through custom tokenization of 8000+ trading-specific terms.
- Dynamic Model Handling: Developed robust real-time query resolution system with dynamic model handling for improved accuracy and performance.
Outcome:
Delivered a high-performance AI assistant capable of processing complex trading concepts with high accuracy and minimal latency.
AWS Infrastructure Automation
AWS
Terraform
GitHub
Automated infrastructure provisioning using Terraform, GitHub Actions, and Packer, integrating AWS services for scalable deployments.
Key Metric: Reduced deployment time by 50%.
AWS Infrastructure Automation
Duration: Sept 2024 – Nov 2024
Technologies: AWS, Terraform, Packer, GitHub Actions, Node.js, SQL
Overview:
Developed a fully automated AWS infrastructure solution that streamlined deployments, enhanced scalability, and ensured robust system performance.
Key Contributions:
- API & Service Integration: Developed RESTful APIs with Node.js, integrating AWS services (SNS, S3, RDS) & implemented Lambda for real-time notifications.
- Infrastructure Automation: Automated provisioning using Terraform for VPC, EC2, RDS, Auto Scaling Groups, ALB, and security groups.
- CI/CD Pipeline Development: Built automated pipelines with GitHub Actions and Packer for custom AMIs, resulting in faster, more reliable deployments.
- Monitoring & Diagnostics: Enabled comprehensive monitoring with CloudWatch, custom metrics, and centralized logging for quick diagnostics.
- Scalability & Reliability: Leveraged Auto Scaling Groups, Load Balancers, and Route 53 for seamless scaling under load.
Outcome:
Delivered a scalable, reliable, and fully automated AWS infrastructure that significantly reduced manual intervention, accelerated deployment cycles, and enhanced overall system performance.