Aug 2024 – Mar 2025 · 7 mos · College Park, Maryland, United States

• Worked on efficient architectures in LLMs and MLLMs under Prof. Tianyi Zhou. Researched on efficient MOE architectures, developed better and more optimal expert training strategies.

Jul 2023 – Aug 2024 · 1 yr 1 mo · Bangalore Urban · On-site

• Pioneered the innovation of India’s first LLM Chatbot for Indic languages as a founding member, leading research across MoE, Multimodal Systems, Large Language Models (LLMs) like ELMo, BERT, OpenAI GPT-2/3, BART, BigBird, ERNIE.
• Automated 60% of customer interactions by outperforming LLama2 Chat SFT on benchmarks such as COPA, ARC and HellaSwag, through optimization and fine-tuning using SFT, Peft-LoRA, DPO, and CPT.
• Engineered the Orchestrator pipeline and Web-RAG architecture for the Krutrim Chatbot. Enhanced the Wav2Lip model for multilingual lip-syncing, increasing resolution by 12%, handled the audio diarization and identification pipeline.

Jun 2022 – Jul 2023 · 1 yr 1 mo · Bengaluru

• Collaborated in reinforcement learning research, boosting driver allocation efficiency and generating a 5% increase in revenue.
• Invented Proximity Unlock technology for OLA e-Scooters with a 93.7% accuracy using on Transformers and LSTMs.
• Developed a Criticality Score prediction model with 90+% accuracy to prioritize and resolve high-impact customer issues, and deployed a State-of-Health (SoH) prediction model for e-scooter batteries, achieving 92.7% accuracy.

Feb 2022 – Jun 2022 · 4 mos · Bangalore Urban, Karnataka, India · On-site

• Worked as a Machine Learning engineer in the advertisement team for building models to predict when advertiser will go out of balance based on their history, also worked on the deployment of the model on AWS and analyse the data flow.

Dec 2020 – Mar 2021 · 3 mos

• Worked as a Data Science engineer, responsible for automation and deployment of various recommendation and venue ranking systems on the company site, analyzing the data, making user-specific changes on the site.

Nov 2020 – Feb 2021 · 3 mos

• Co-Founded a Traffic Count and Tracking venture, implemented advanced object detection algorithms for traffic monitoring, optimizing vehicle tracking within the application; achieved 93% accuracy rate in identifying vehicle types and toll collection data.

Oct 2020 – Dec 2020 · 2 mos · Vancouver, British Columbia, Canada

• Worked in the ML team for building and reviewing recommendation models, using various ML approaches and compiling them.
• Worked as the team lead in Data collection and scraping, for hiking destination sites and booking hotel sites, cleaning and analyzing datasets, and as a Team-lead for Recommendation models based on clustering for restaurant recommendation.

Aug 2020 – Dec 2021 · 1 yr 4 mos

• Led one of the most renowned Teach-Team of DTU, in the field of AI and Robotics. Worked with a team of aspiring students of DTU in building bots that are autonomous, working on participating in international competitions like IGVC-2021, and also hackathons for bot simulations and research in the same.

Aug 2019 – Aug 2020 · 1 yr

• Worked as the Head of Navigation Dept. of UGV, a Tech Team in Delhi College of Engineering.
• Worked on a project of creating a self driving, autonomous unmanned ground vehicle, using Machine Vision and Deep learning concepts, to participate in IGVC-2020, built navigation stacks for the bot, experimented and incorporated various ML and computer Vision methods for auto-navigation tasks and simulations and taught juniors for the same.

Feb 2019 – Aug 2019 · 6 mos

• Worked under seniors of the Tech-Team and learnt methods of Computer Vision for autonomous vehicles, ROS platform and frameworks and started contributing to bots and navigation stacks.

May 2020 – Jul 2020 · 2 mos · Delhi, India

• Worked as a Research Intern, building projects and processing datasets for issues and real life problems like Antiviral-drug prediction for Covid-19, with a group of other students.

Jan 2020 – Apr 2020 · 3 mos · Work from home

• Worked as a Research Intern for implementing and improving ML and Deep Learning solutions and techniques on real life problems and projects based on them.
• Worked on building a NLP based LSTM Model for analysing the Reviews on a particular product provided on Youtube by consumers, compare these views and generate a score based on the type of reviews for that product, after scraping the transcripts of the videos via Youtube API.
• Developed an AI model using Deep Learning techniques like SVM Segmentation and YOLO Object Detection to detect and classify the type of vehicle in a on road camera video and predict the direction of motion for the vehicle.

Aug 2018 – Jan 2019 · 5 mos · DCE Delhi College of Engineering, Delhi

• Worked for creating the image processing codes for the Rover, building software networking using ROS and making the project autonomous.