# Github Information:
Hi 👋 My name is Siddhant Rai Research Engineer - ML @ Siemens I am an Artificial Intelligence and deep learning enthusiast. I love to explore new things and learn about them. I am mesmerized by the amount of audacity Artificial Intelligence holds to empower our lives. I am into competitive coding too, I am currently in the top <contact_num> on Hacker Rank worldwide. I love to code things from scratch, from making a Multi-layered Artificial Neural Network to writing my Genetic Algorithms to play a game like pong, flappy bird, etc. I invite you to this world of AI that could be the next Pandora, are you ready to embrace yourself as the next Pandoran? 🌍 I'm based in India ✉️ You can contact me at <email> 🧠 I'm exploring Uncertainty in Machine learning and ways to effectively handle it. Skills Socials Badges My GitHub Stats# Resume: 
Siddhant Rai <contact_num> | <email> | github.com/siddhant230 | linkedin.com/in/siddhant-rai/ |Â rsiddhant73 Problem solver and Engineer Highly skilled in core Deep learning, Computer vision, NLP , GenAI, multi-modal architectures, Self- supervised learning (Low resource/data regime tasks) and optimization EXPERIENCE Siemens, Lead Research Professional (Vision, NLP & optimization) (Bengaluru, India) Aug <contact_num> - present • Developed GPT4-Vision based sensor extraction and characterization module from schematic diagrams. • AiDAMA : Developed end-to-end information extraction pipeline using detection and custom one-shot-classification systems for P&ID diagram digitaliztion, wrote Synthetic diagram generation algorithm to generate usable and trainable dataset. • AI4Safety : Developed, deployed and field tested the Computer vision based Survelliance and alert mechanim., currently being used in multiple Siemens factories . • E-mobility : Pioneered time-series forecasting and trajectory modeling for efficient vehicle routing, while also devel- oping interactive dashboards for anomaly detection and trend visualization . • Core member in Foundation models team for VQI and semantic tasks, development and benchmarking of available models. • Part of DeltaQ innovation team ; majorly dealing with replication, enhancement and rule space maintainence of existing VQI platform • Auto annotation tool for zero and one shot image annotation , extensively used in Ai4Safety and other internal projects. • Procedural Video generation : Mechanism to generate sequential industry level instruction video using text as prior, pipeline included models such as SDXl and SDXl-turbo with controlNets for end-to-end generation and video editing. • Optimization for EV route planning and charge management using Quantum inspired Computing . SpringW orks, Machine learning Intern | Engineer (Vision, NLP) (Remote) (Bengaluru, India) May <contact_num> - May <contact_num> • Conducted satellite image segmentation and one-shot retrieval using custom designed architecture and training meth- ods to assess lawn area, optimizing cost estimates for mowing and plowing services. The developed solution improved upon the previous by around 40% reaching accuracy of 98%. • Developed custom architectures and backend systems to extract entities from government documents , enabling seam- less e-verification for Aadhar card, P AN card, and more. • Established a face matching service by integrating document proofs and applicant selfies, enhancing identity verifi- cation processes . • Developed a resume parsing service , employing rule-based and custom LSTM architecture with attention mech- anisms to extract essential information from resumes and CV s, enhancing information retrieval efficiency , the developed method was 10x faster than previously employed process . • W orked on the creation of a Job recommendation engine , leveraging resume data and previous performance metrics to optimize job matches with job descriptions, enhancing candidate placement accuracy . • Developed procedure to generate multi-choice QnA using rule sets and advanced techniques including GPT-3 and prompt engineering. • Designed in-house profanity filter for filtering out discriminating question-answers. Astute Resolutions, Machine learning Intern (Vision) (Remote) (Lucknow, India) Aug <contact_num> - Oct <contact_num> • Developed surveillance algorithms , including camera temperature detection and fire/smoke detection, enhancing safety measures using custom designed light-weight CNNs . • Spearheaded a parallel parking project, configuring both backend and UI components. Devised an algorithm for optimal vehicle positioning within tight parking spaces. MENTORSHIP HealthFlex, T echnical mentor and advisor (optimzation) (Bengaluru, India) Jan <contact_num> - Present • Mentoring and developing 9DoF IMU driven pose estimation and trajectory alignment verification for a given move- ment/exercise. • Developed core sensor signal parsing and conversion module • Helping interns to develop rules and patient onboarding protocols. Springworks, T echnical mentor (Vision & NLP) (Bengaluru, India) Jan <contact_num> - Mar <contact_num> • Mentored and helped the internal team to develop RAG based slack chatbot (Albus) utilizing Langchain and LlamaIndex as well as vector databases . • Helped in prompt optimization and design as well as code quality checks. • T ackled various problems related to response management and profanity checks. Springworks, T echnical mentor (Vision & NLP) (Bengaluru, India) Nov <contact_num> - Jan <contact_num> • Mentored interns, ensuring code quality and offering technical support to foster their growth. • Delivered comprehensive explanations of various deep learning algorithms, building a strong foundation from the ground up.SKILLS Languages Python, C/C++, Java Framework T ensorflow, Pytorch, JAX, HugggingF ace, RAG, LangChain, LlamaIndex, Flask, F alcon, Streamlit, Gradio Software Linux, Docker, OpenCV, Git, Bash PROJECTS Self-conditioned Diffusion model for Time series modelling and interpolation F eb <contact_num> • Developed custom Score based Diffusion model for approximating temporal state dynamics. • The model follows a Non-Equilibrium Time Dynamical system to sample noise across time. • Main task was to model trajectory of time-series data with missing data points and approximate them by temporal interpolation . • The algorithm further used to understand and quantify data complexity by computing Jacobian of energy surface (basically a gradient map). Visual Question Answering for scene analysis Dec <contact_num> • Utilized open-sourced BLIP2, groundingDiNo and SAM to create framework for scene understanding, localization and retrieval . • F or more detailed and structured outputs, replicated same setup with Gemini-pro (vision) instead of BLIP2. Crypto clustering and trajectory analysis Dec <contact_num> • Implemented algorithms to find correlation between different cryptos and cluster them together using DTW . • Idea was to analyze the major cryptos and their associated sub-cryptos , such that by analyzing few parent coins we can largely predict other sub-crypto behaviours .. • The project and algorithm was eventually used by SpringW orks for internal purposes and partnerships. Vision transformer Jan <contact_num> • Implemented original vision transformer from scratch including the attention blocks and patching layers . • Idea was to have custom architecture for low resource downstream tasks. Neuro - Evolutionary Algorithms for Games Jan <contact_num> • The main idea of the project was to augment famous games like Flappy Bird,Pong and Space Invaders and automate them using concept of Evolutionary Learning. • Implemented Multi-layered Neural Net from scratch using Numpy along with Back-propagation algorithm. • Implemented a custom weight updation module inspired from randomised weighted gradients/weighted mutation. INVENTIONS AND PUBLICATIONS •Submitted 13 Invention disclosure on topics ranging from VQI on digital twins to Encryption and data compression. •Paper on Object detection and zero-shot retrieval of symbols from P&ID diagrams (under review). •Uncertainty estimation in models for structure predictions (in progress). COLABORATIONS and COMMUNITY Genesis AI Labs, Research Collaborator F eb <contact_num> - present • Research on Protein structure prediction, drug discovery and other Geometric DL works. CORD.ai, Research Collaborator Jul <contact_num> - present • Research collaboration on Computer vision and NLP tasks. EDUCATION Lovely Professional University ,B.T ech in Computer Science | Punjab, India GP A: 8.5 / <contact_num> City Montessori school ,Class 12th | ISC | Computer Science | Lucknow, India percent: 94% <contact_num> City Montessori school ,Class 10th | ICSE | Computer Science | Lucknow, India percent: <contact_num>% <contact_num> INTERESTS T echnical WritingI write descriptive blogs about topics ranging from Protein structure prediction to Generative models and core optimization some of them are on my LinkedIn and Medium . Art I have always enjoyed drawing and sketching, mostly I do landscapes. Puzzles I love solving puzzles, riddles and cubes Books Cherish reading about philosophy (both west and east) as well as fiction books. Games Quite skilled in track events like 100, 200m sprints as well as sports such as football and badminton.