Praveen Kumar P
Available Thursday
Starting at 3:30 AM UTC
Questions?
You can start chatting with Praveen Kumar before you get started.
Work with Praveen Kumar
Select a category that you're looking for help in.
Work Experience
Applied AI
JPMorgan Chase & Co.
February 2025 - Present
I hold the Vice President designation (senior banking band), yet my focus remains hands-on architecture and engineering. Spearheading personalized customer experiences and actionable insights by leveraging Explainable AI, large-scale financial data, and bleeding-edge cloud technologies. ** IC: Individual Contribution 1. [IC] Transaction Search using Embedding and Small Language Models (SLMs): Designed and implemented an intelligent transaction search system capable of handling complex, multi-intent user queries using embeddings and SLMs. 2. [IC] Budget Pacing Optimization: A data-driven, mathematical optimization framework combining machine learning and mathematical optimization to maximize ROAS and maintain balanced campaign spend. 3. [Presenter] Agentic AI Training Initiative (Brown-Bag Series) to the broader-team in JPMC: Deliver weekly 1-hour training sessions (to broader group) focused on hands-on experimentation and conceptual deep-dives in Agentic AI. Initiative reflects continuous learning and leadership in next-gen AI systems. Giving Back to the AI Community: Deliver public talks and invited sessions on Agentic AI as part of community contribution. Have spoken at IIM Ahmedabad, IIIT Hyderabad, and other leading institutions on building reasoning-driven, enterprise-scale AI systems. Skills: Agentic AI 🔹 Generative AI 🔹 Computer Vision 🔹 Large Language Models (LLMs) 🔹 NLP 🔹 Large Multimodal Models 🔹 Team Leadership 🔹 Team Management
Staff Scientist
krutrim.com
February 2024 - January 2025
Multi-modal LLM team: Focus is on product driven research based on Large Multimodal Models (LMMs) and Diffusion Models. Skills: Deep Learning 🔹 Generative AI 🔹 Computer Vision 🔹 Large Language Models (LLMs) 🔹 NLP 🔹 Large Multimodal Models 🔹 Team Leadership 🔹 Team Management Published Papers: - Chitranuvad: Adapting Multi-Lingual LLMs for Multimodal Translation (https://aclanthology.org/2024.wmt-1.80.pdf) - Chitrarth: Bridging Vision and Language for a Billion People (https://scholar.google.com/citations? view_op=view_citation&hl=en&user=tgz2F38AAAAJ&sortby=pubdate&authuser=1&citation_for_view=tgz2F38AAAAJ:7PzlFSSx8tAC) Contributions: [IC: Individual Contribution, Collaborative: Joint work with team] 🔹 Image Moderation System (IC): Developed image moderation system (Data collection & model development, model evaluation), ensuring robust handling of harmful content for vision-language models (VLMs) and diffusion models. 🔹 Indic VLM Eval Datasets (Collaborative): Leading initiatives to curate diverse benchmarking datasets that captures linguistic and visual nuances specific to the Indian context for accurate benchmarking of VLMs. 🔹 Indic VLM Evaluation Pipelines (IC): Spearheading the creation of an evaluation code pipeline, enabling precise benchmarking of VLMs across various capabilities and skills. 🔹 Research Publications (Collaborative): Driven by a passion for advancing research at the intersection of computer vision and large language models, I focus on innovating in multimodal AI, contributing to co-authored publications on Large Multimodal Models (LMMs). 🔹 VLM Finetuning (IC): Created a fine-tuning coding pipeline tailored for proprietary OCR datasets.
AI INstructor
upGrad
February 2023 - January 2024
Taught Data Science and Computer Vision courses to the masters program offered by UpGrad. Skills: 🔹 Teaching | 🔹 Data Science | 🔹 Computer Vision | 🔹 Curriculum Development | 🔹 Education Technology 🔹 Python for Data Science | 🔹 Machine Learning | 🔹 Academic Mentorship | 🔹 Online Learning Platforms | 🔹 Instructional Design
Sr. Manager, Data Science
jioplatformslimitedjpl.com
June 2022 - February 2024
Topic of interest: Video Analytics, Computer Vision Group Skills: Deep Learning 🔹 Generative AI 🔹 Computer Vision 🔹 Large Language Models (LLMs) 🔹 NLP 🔹 Large Multimodal Models 🔹 Team Leadership 🔹 Team Management While at JPL, I was engaged in two projects. The first project involved the development of video-based foreground segmentation and background-aware foreground segmentation techniques. The second project focused on context-driven object detection. As a result of these endeavors: Our research manuscript that include innovative architectures for video-based foreground segmentation is available online (URL: https://arxiv.org/pdf/2402.00918.pdf), demonstrating my commitment to advancing the field of computer vision. I created a custom repository utilizing Ultralytics-YoLov8 with PyTorch for the second project, showcasing my proficiency in implementing and enhancing object detection models. I collaborated on object detection influenced by context, and additionally, I worked on Optical-flow prediction using Global Motion Aggregation (GMA).
Lead Engineer
Qualcomm
September 2021 - May 2022
Skills: Deep Learning 🔷 Computer Vision 🔷 Team Management During my tenure at Qualcomm, I undertook and successfully executed two key projects: I. Developed a Face Detection model tailored for smartphone cameras, adhering to specific criteria: Engineered a compact and quantized deep learning model with low power consumption and latency, specifically optimized for screen-off scenarios. Achieved superior detection performance while minimizing false alarms, particularly on challenging datasets. Implemented a regularization strategy leveraging both face landmarks and face orientations to enhance the stability of the compact model. II. Contributed to the creation of a Gaze simulator aimed at generating a larger dataset with annotations. This initiative was part of the ongoing efforts to develop a compact Gaze detection system. Throughout these projects, the environment utilized was PyTorch. PyTorch, Python (Programming Language) and +6 skills
Education
Indian Institute of Science
Doctor of Philosophy - PhD, Artificial Intelligence, Machine Learning, Mathematical Optimization, and Computational Imaging
2015 - 2021
Best PhD Thesis Award & Gold Medal(Prof. D J Badkas Medal) TensorFlow, Statistical Modeling and +4 skills
Indian Institute of Technology, Guwahati
M.Tech, Signal Processing
2007 - 2009
Grade: 8.54/10
Select a category you’re looking for help with:
No posts yet
When Praveen Kumar posts, it will show up here.