Smita Singh

- End-to-End Development of Documentation QA Bot: Built a question-answering bot using the RAG mechanism, enabling efficient and accurate document-based query resolution. - Accounting Task Automation with LLMs: Developed models to address accounting challenges, such as identifying the correct accounting table for bank statement items. - Data Analysis: Analysed and interpreted accounting data to derive insights and support model development. - Accounting Chatbot Development: Worked on LLM-powered chatbot tailored for accounting use cases, improving accessibility and efficiency in financial data management. - Proof of Concept (POC) for Memory Component: Prototyped a memory module for the chatbot, inspired by the MemGPT architecture, to explore possibilities for enhancing conversational context retention and user experience.

- Throughput Optimization: Applied reinforcement learning techniques to optimize the semiconductor manufacturing process, improving production efficiency and throughput. - Model Development: Utilized Python, TensorFlow, and Rllib to design and train models within the OpenAI Gym environment for simulation and optimization. Tech Stack: Python, TensorFlow, Rllib, OpenAI Gym

- Time-Series Data Processing: Analyzed and processed time-series data containing battery performance and lifecycle information. - LSTM Model Development: Designed and implemented a Long Short-Term Memory (LSTM) model to predict battery lifespan with improved accuracy.

- Command Auto-Completion: Built a Linux command auto-completion feature using a “tab-tab” shortcut for faster bot data submission. - Device Management CLI: Developed a tool to fetch device status, identifying locked and free devices. - Optimization: Enhanced auto-completion efficiency and reduced response time. - Structured Output: Improved the readability and organization of auto-completion results.

- Contributed to the SmartThings ecosystem by enhancing the Smart AC hybrid application - Developed a hybrid mobile app for Smart AC, enabling seamless control of AC settings and improving user accessibility across platforms. - Upgraded legacy code by migrating the Smart AC app to the latest Angular JS version, enhancing performance and maintainability. - Supported testing initiatives, ensuring robust functionality and reliability for production deployment. Technologies Used: Angular JS, HTML, CSS, C++, SQL

Freshsales - A CRM software - Integrated third-party apps (e.g., Xero) via OAuth, enabling sales teams to access customer invoice data for improved relationship management and lead generation. - Designed and implemented pagination, optimizing invoice visibility for sales reps to navigate large datasets efficiently. - Developed filtering capabilities, allowing users to sort invoices by status (paid, unpaid, open, closed) and link/unlink records, boosting operational flexibility. Tools and technologies used: Ruby, Rails, MySQL