Shaurya Dwivedi Email & Phone Number

Malmö, Se

Malmo, Skåne County, Sweden

Bengaluru, Karnataka, India

-Fine-tuned FLAN-T5 model to adapt to the telemetry data to create a dialogue summary.-Implemented Performance efficient fine tuning(PeFT) to optimize the FLAN-T5 model.-Implemented Low-Rank Decomposition(LoRA) to reduce the trainable parameters, improving memory efficiency andfine-tuning speed.-Implemented Reinforcement Learning Human Feedback mechanism to fine-tune the FLAN-T5 model.-Improved FLAN-T5 model by aligning them with complex human values through RLHF by training the reward model.-Enhanced LLM-generated responses by combining retrieval-based(RAG) and generative approaches.-Utilized Distillation techniques to transfer knowledge from a large, complex model (teacher) to a smaller, more efficientmodel (student).-Quantized precision of model weights and activations from 32-bit floating point to 8-bit integers.-Deployed Llama3.1 ,by compressing it to INT4, on Intel edge devices using OpenVINO.-Led a team of 7 developers.

Bengaluru, Karnataka, India

*Friction Less Checkout- Automated complete shopping experience of a typical grocery store by reducing 90 per cent human intervention.- Architected the end-to-end pipeline for technical implementation.- Developed inference module for Human Pose Estimation using the Open Pose approach with OpenVino as framework.- Trained 3D-Resnet-50 model for Action Recognition with a confidence score of 76 per cent.- Converted the trained model to onnx and quantised to fp16 and fp8 precision for edge inferencing.- Developed inference module for 3D-Resnet-50 at 18 inferences per sec for upcoming server architecture.- Implemented KNN-based background subtraction using OpenCV.- Trained Resnet-18 model for Product Classification(38 classes) with a confidence score of 98 per cent.- Developed inference module for Resnet-18 at 45 inference per sec for upcoming server architecture.- Led a team of 3 developers.

Bengaluru Area, India

*Driver Monitoring System(DMS)- Responsible for developing Hand on Face Detection Module.- Developed Hand on Face Detection Module using tensorflow and deployed on soc using tflite.* Surround View Monioring System( SVM)- Responsible for developing and deploying SOTA Object Detection Models on SoCs(Qualcomm 8155).- Lane detection on Fisheye images.- Empty parking slot detection.- 360-degree Surround View Monitoring System with Fisheye images- Image sticthing using state-of-the-art Computer Vision algorithms.- Sensor fusion with Steering wheel guidelines and Ultrasonic sensor based alerts.- Our Demo at CES 2020https://www.youtube.com/watch?v=Bawur9yGMuU&t=147s

Pune, Maharashtra, India

* Converted Muliple client POCs to project- Developed HRSG baffles inspection using Object Detection techniques for quick scan of defects. (Patent Applied)- Increased inspection 20X by automating the inspection task.- Extensively used Python and Tensorflow for the implementation.- Developed end to end pipeline for Audio Classification using Deep Learning- Converted audio frequency data to images representing the spectrogram features.

Mysore

* Self Driving Car - https://www.bloombergquint.com/technology/infosys-building-self-driving-golf-carts-to-be-future-ready- Selected among top 100 programmers at Infosys via Hackathon organised by Hacker Rank.- Used Deep Learning for Computer Vision for Object Detection, Traffic Sign Classification and Semantic Segmentation.- Used Kalman Filters and Particle filters for Sensor Fusion and Localization.

Pune, India

* World Anti Doping Agency(WADA)- I was responsible for writing code in JAVA for the legacy shift of the existing application.- I used JSF 2.2 with HTML and CSS for the front end.- Spring Boot, Hibernate, SQL and JMS servers were used for the backend of the application.

Mysore

*Joined as a systems engineer trainee- Developed Desktop application using Python for banking as a project.- Trained in OOP using Python- Trained in OOP using JAVA- Trained in Front End Technologies such as HTML, CSS, Bootstrap- Trained SQL

Master Of Technology - Mtech, Data Science, A

Activities and Societies: CSE Department Social Media Outreach Admin* Courses - CS5590 Foundations of Machine Learning - CS6660.

Self Driving Car Nano Degree, Artificial Intelligence

Studied machine learning, deep learning, computer vision, sensor fusion, robotics.

Bachelor Of Technology (B.Tech.), Computer Science And Engineering