- Spearheaded the development and optimization of graph-based Retrieval-Augmented Generation (RAG) techniques for Large Language Models (LLMs), enhancing the accuracy and efficiency of information retrieval processes.- Engineered and integrated Neo4J knowledge graphs with LLMs, enabling advanced semantic understanding and context-aware responses in generative AI applications.- Designed and implemented hybrid LLM architectures, combining diverse AI models to leverage their strengths and improve overall system performance.- Led research initiatives in generative AI, contributing to cutting-edge development in natural language processing and machine learning.- Collaborated with cross-functional teams to translate complex technical concepts into actionable solutions, driving impactful outcomes for various AI-driven projects.

Ericsson ONE is an internal incubator for innovations and startup ideas.

>> Addressed complex business problems through Statistical Learning methods that include deep learning based computer vision and text analytics techniques, end-to-end Machine Learning workflow design, data wrangling, data cleaning, feature engineering, development of robust Supervised and Unsupervised Machine Learning algorithms, optimization and deployment of predictive models as Flask APIs or in docker. >> Worked extensively on clustering, classification, regression, predictive modeling on big data using Spark (Scala and Python - PySpark), multivariate analysis, statistical analysis using IoT, LTE, 4G and 5G data. Expert in 5G core network domain, 5G Service Based Architecture, IoT, Edge and Fog computing .>> Drones capture movement of traffic images at different locations of a city and images are sent in stream to edge servers. Developed CNN based Deep Neural Network to detect high traffic at edge servers. Consolidated result is sent to Cloud to take further decisions and alarm generation.>> Developed parser for extracting user events from moderate to massive, semi-structured text data coming from highly distributed, third party network adaptors.Developed NLP and Machine Learning algorithms for analyzing and predicting performance degradation from user event data. Deployed the model in production.>> Developed Machine Learning models to predict SLA breach of support tickets that reduced actual SLA breach by 35%.>> Developed nine patent-pending solutions as individual inventor. >> End-to- end designing and implementation of statistical modelling, advanced data science solutions and products from acquiring business needs to final delivery, developing predictive models for field service operations for major telecom operators, predicting faults in LTE networks and new product demand forecasting.

>> Video Analytics with Hadoop development - Development of proprietary tools, syndicated services and custom analytics to provide actionable insights to differentiate product, site or brand, to target audience, to define the business objectives and to power online strategy.Video Analytics provides daily/weekly/monthly insights into viewers’ population, viewers’ interaction with video, video advertising, end to end development of new stable, scalable features.>> Contributed in simulation modeling from text logs to provide daily/weekly/monthly insights into viewers’ population, viewers’ interaction, video advertising and end to end development of stable, scalable features that helped client to increase the targeted traffic by 23% at initial level.>> Technical and strategic decision making on critical incidents by alternative analysis, predictive analytics, estimation modelling, data analytics and simulation modelling.>> Performed predictive analytics, performance tuning and scheduling of stored procedures and functions in MS SQL production server for high velocity manufacturing data that eliminated hour long slowness experienced during pick hours.>> Development of LIMP (Laboratory Information Management Systems) for TCS Centre of Excellence (Technology: asp dot net, MS SQL 2008, MES) - Developed a new web based LIMP product from scratch for TCS CoE.

I designed and developed a heuristic algorithm to place components on PCB layout. I designed the layout first by KiCAD, an open source PCB design software from CERN, that generates netlist and PCB files. My placer reads these two files and runs the algorithm to place components in optimal way and to reduce the total wire length.

**IPTV middleware development and analysis**>> Developed auxiliary component for existing BSS. >> Performed data analysis on video broadcasting data to extract subscribers’ VoD (Video on Demand) and SVoD (Subscription Video on Demand) buying pattern.

>> Taught Electronics Devices and Circuits for one complete semester to students of Electrical Engineering, second year.>> Was assigned to take the labs for the students of Electrical Engineering, second year.>> Arranged Stuff Development Programme for the Department of Electronics Engineering, PCE.>> Was assigned as the warden of Priyadarshini Boys Hostel having capacity of 300 students for one complete semester.