• Developed and implemented automated solutions for customer support, leveraging a RAG-based agentic framework and function similarity with attachment hierarchies to significantly reduce manual effort and task resolution time

• Fine-tuned various LLMs (Qwen, GPT-OSS, Mistral) using supervised fine-tuning to improve response quality for Excel and text inputs using unsloth. Developed an efficient inference framework with vllm.

• Utilized langgraph to create agentic framework for multi-comment support tickets helping in understanding which attachment to use and how to merge the attachments to resolve customer support tickets.

• Successfully processed varied unstructured data (Text, Images, Excel, PDF), efficiently extracting essential information to accelerate the resolution of targeted customer support issues.

• Utilized PGvector with L2 similarity search and keyword search to accurately analyze customer messages, retrieve relevant information, and effectively solve customer support issues

• Pioneered initiatives in customer support automation, focusing on reducing time spent on mundane tasks and improving overall efficiency by intelligently utilizing information retrieval and agentic systems

• Contributed to a test-driven development (TDD) environment, ensuring over 93% test coverage.

• Collaborated on a project that systematically queried diverse organizational data within decentralized environments.

• Utilized Terraform, Golang, and Gorm to develop functionalities for managing access permissions and facilitating seamless uploading of large files.

• I applied AWS Glue using Terraform to convert diverse file data into a table format.

• Conducted Load Testing of applications using JMeter and Blazemeter for performance evaluation.

• Led the Research and Development efforts to implement a Reinforcement Learning-based solution for optimizing crop drying processes.

• Designed and developed datasets for crop drying procedures, conducting thorough analysis to drive improvements.

• Explored blockchain-based solutions for transaction monitoring, collaborating with Hyperledger BESU and Fabric teams and initiating a private blockchain platform with Blockapps.

• Enhanced C# and Python code to optimize corn drying procedures, improving overall effectiveness.

• Led and mentored a cross-functional team of developers, testers, and business analysts to deliver innovative blockchain projects. Technologies included Angular 8, Node.js, Solidity, React Native, AWS, and Azure for AI-based APIs.

• Engineered an OAuth2-based login system using Amazon Cognito and integrated Federated login via Facebook within Angular 8.

• Implemented og-tags for generating shareable URLs on social media platforms using Pug templates in Node.js, enabling server-side rendering.

• Designed and developed a financial investment and crowd-funding social platform integrated with a payment gateway for seamless conversion of USD to blockchain coins (ERC20 Tokens).

• Established a serverless architecture using AWS Lambda for a Blockchain-based REST API application, ensuring scalable and efficient connectivity.

• Conceptualized, implemented, and assessed novel models and rapid software prototypes in machine learning to address diverse challenges.

• Applied advanced models, including LSTM, GRU, and Elastic search, to develop a marketing optimization solution, predicting promotions based on historical data.

• Engineered feature sets for model creation, evaluating gradient boosting techniques employing Light GBM and CatBoost models.

• Conducted in-depth analysis of customer behavior, leveraging categorical, continuous, and text analysis to predict optimal promotions and evaluated model performance using metrics such as F1 score and accuracy.

• Employed minhash-lsh and xgboost algorithms for deduplication tasks, generating shingles from text data and calculating minimum edit distances for efficient record matching.

• Devised and implemented deduplication and search algorithms utilizing fuzzy search in Elastic search and xgboost models, optimizing search engine performance.

• Leveraged libraries such as pandas, numpy, pytorch, and nltk to enhance search accuracy and efficiency, ensuring robust outcomes.