Principal Software Engineer
Qualcomm
مجموع سنوات الخبرة :13 years, 11 أشهر
● Implemented Microservices oriented ML-based PDF Parsing using FastAPI, Postgres, SQS and S3. Deployed on EKS.
● Implemented KEDA deployment for the ML solutions on EKS.
● Worked extensively on time-series based problems. Implemented projects that involved curve fitting, Arima, time series models for finance, sales and demand forecasting, deep learning for time series, hierarchical time series, hybrid method for time series and transfer learning for time series.
● Built a real-time system using to perform analytics on streaming data as it is generated for dynamic pricing strategy
● Used Kafka to build an even driven system for training and re-training a time series machine learning model to predict prices of securities and options.
● Leading a team that designed and built a high through-put, low latency Real-Time Data processing platform using AWS Lambda, Kinesis, S3, Amazon DynamoDB, Amazon Cognito and Amazon Athena
● Built an AI service platform using Microservices architecture on Kubernetes (EKS).
● Fine Tuning Transformer models on Domain specific data for generating custom dense vector embeddings.
● Built end to end ML operations pipeline to develop, train and deploy the model on cloud. Implemented the backend for serving ML model using distributed micro-services using FastAPI, RabbitMQ.
● Designed and implemented high through Data Ingestion pipeline using Fastapi, Redis Celery and Postgres.
● Implemented the complete SDLC using CI/CD pipeline on Gitlab.
as a software engineer for developing backend web services and implementing machine learning using Python.
•Designed the backend server for the software bug monitoring tool
•Created Restful API for integrating machine learning software into the application.
I developed software systems for Machine learning/Artificial Intelligence and Web Development using Python, Java.
Deep Learning and Artificial Intelligence
• Implemented simulation software in Java using extensive knowledge of advanced data structures and algorithms.
• Built a face recognition system in Java using machine learning libraries.
• Built Machine learning systems using convolutional networks, RNNs, LSTM, Adam, Dropout, BatchNorm, Xavier/He initialization using python and tensor flow.
• Built state of the art healthcare, autonomous driving, sign language reading, music generation, and natural language processing applications using deep learning models in Java and Python.
• Built software application for improving advertisement efficiency for an ecommerce company using deep learning.
• Built a face recognition software system using machine learning for a home automation start-up company.
• Designed and built speech recognition and Machine Translation software using Python.
• Designed NLP applications using Recurrent Neural Networks models and other time series models like Markov Models.
• Chatbot with TensorFlow's sequence to sequence library and by building a massive database from Reddit comments.
• Created Machine learning algorithm for predicting stock market prices.
• Created software for Halite II 2017 Artificial Intelligence Competition.
Web Development
• Web Scraping
• Crawl web sites and pull information from online sources.
• Designed advanced and complicated e-commerce web applications using Angular 2.
• Designed web application using Angular 2 and frameworks like firebase, MongoDB.
• Implementing authentication and authorization using Auth0 and JSON Web Tokens (JWT).
• Using Reactive Extensions and observables.
• Optimize the rendering performance of your Angular apps.
Started a software company to develop algorithm trading software for banks and hedge funds.
• Designed and built a low latency, real-time market data ticker plant and trading platform using C/C++, java Multithreading.
• Designed and implemented real time data feeds using exchange specific raw protocols and Apis, Ex: BSE, NSE etc.
• Developed a pub-sub model to publish normalized proprietary messages on multi-cast and tcp/ip adapters
developed internally.
• Developed a framework to modularly place proprietary algorithms inside the platform to adapt to various business objectives.
• Developed robust, low-latency, real-time data consuming client Apis for c and java users.
Platform:RHEL 6
Languages: C/C++, Core Java
Protocols: TCP/IP, UDP
3rd Party Libraries: ACE
on the target level build and integration of LCU 3.0 PL. This Target is currently in the Post CS phase(already commercialized). Understanding the build architecture and writing make files to automate the build procedure.
This involved writing makefiles to compile the various modules of the software running on the mobile chip. The task was to make sure that the compilation was not broken and to fix the compilation errors if they existed.
•Testing of the physical level functionality of the software by downloading it on the mobile chipset.Various basic functionality like calling, messaging, data calls had to be tested. Scripts written using Perl and Python were used to write test cases to run on the software.
•ARM processor debugger JTAG was used to fix crashes on live running mobile chipsets. Software debugging using the JTAG debugger formed a core part of my daily job. In the event of the mobile crashing, JTAG was used to analyze the crash and identify the area to be fixed. Also any broken functionality was fixed in the similar manner.
•Power dashboard testing using propriety Qualcomm hardware and software was also part of my job. This involved testing current consumption for various physical level functionality of the mobile chipset like voice calls, data calls, messaging, sleep current and stand-by current consumptions etc.
•Modeling of Network using mathematical programming models. Optimization using integer programming and non integer programming using tools such as CPLEX, Gurobi, AMPL was carried out.
Network coding techniques for broadcast/multicast 4G Networks to improve retransmission capabilities and increase network performance.
•Designed retransmission schemes for LTE Networks which can be implemented at the base station.
•These techniques can be part of the standard HARQ implementations in the future standards LTE, 4G.
•Designed and implemented algorithms for memoryless decoding of these schemes at the UE.
•Worked on different types of transmission schemes for different feedback bandwidths and channel conditions.
•Modeling and simulation of the wireless network with mobile users and their performance analysis.
•Linear programming Modeling and simulation was done with MATLAB and CPLEX. Large scale optimization of network protocols was carried out by using tools such a Gurobi, AMPL etc.
for Wireless Sensor Networks, Proc
Network Coding
Real time Networking embedded systems
•Porting of IPv4 Networking protocols from RTOS platform to Linux
•Platform related development of network security protocols.
•Performance and load testing of internet messaging protocols and standards.
We built an on board pulse code modulation system for the air to air missile Aakaash. We studied the physical layer modulation and demodulation schemes used on the missile and studied methods of equalization.
Publications:
Conferences
), IIIT-