Ojasvita Walgad profile photo 1 on Tal

Ojasvita Walgad

Machine Learning Engineer | AI x storytelling ❤️

Bengaluru

4+ years of work experience

Param Foundation
Sep 20254 mosDec 2025

On-site

An interactive hide-and-seek installation where visitors playfully test ways to outsmart and evade real-time AI detection. Built using Pygame, YOLOv8 ONNX model, and physical sensor systems to spark curiosity around machine perception and pattern recognition.
Jaaga
Mar 20253 mosMay 2025

Bengaluru South, Karnataka, India

I explored the intersection of AI and storytelling—creating a short film using tools like Suno.ai, ElevenLabs, Eden.art, and Gooey.AI. I created Jengai, a short film that was selected for screening at the Dhurva Film Festival, Darjeeling. My work focused on future-positive narratives, training LoRA models, prompt design, and advocating for the ethical use of AI while addressing inherent biases.
hedwigAI
Dec 20235 mosApr 2024

Bengaluru, Karnataka, India

1. Developed a multi-GPU LoRA fine-tuning framework for a Texify vision encoder-decoder, adapting the model to handwritten math equations to achieve a 46.7% reduction in character edit distance and more than double BLEU score (0.21 →0.44) over baseline. 2. Curated a 10k-sample diverse handwritten dataset combining multi-line formulas and algorithms, designing synthetic noise pipelines (elastic distortion, motion blur) and a contrast-enhancement inference step to resolve degradation in real-world phone captures.
RadiusAI, Inc.
Dec 20202 yrs 11 mosOct 2023

Bengaluru, Karnataka, India

Building AI systems for retail stores utilizing advanced computer vision techniques for analyzing customer behavior and managing inventory, successfully deploying them across 700 convenience stores in the US. Experienced with TensorFlow, PyTorch, and deployment frame- works like TensorRT for optimizing and accelerating model infer- ence on NVIDIA GPUs. Proficient in Docker for containerization and Kubernetes (k8s) for container orchestration. Few Projects: 1. Person Re-identification for Tracking in Indoor Cameras - Trained and deployed a Body Part-Based feature learning model for robust people tracking in retail settings. - Achieved significant mAP score improvements: +30.88 points on clients multi-camera data and +12.46 points on public Re-Id dataset (Market 1501), surpassing the Omni-Scale Network (OSNet) re-id model. 2. Multi-Task Network: Monocular Depth and Segmentation - Trained and deployed a multi-task network using a lightweight RefineNet architecture, introducing an enhanced loss function for depth (weighted normal loss and trimmed MAE) and segmentation tasks (weighted Lovasz loss and focal loss). - This model is being used for multiple use-cases: AOI heatmaps (analyzing area-of-interest in retail stores) and 3d interactions (via Voxel Point Clouds). 3. Kalman Tracking and Lens Rectification for Outdoor Cameras - Developed a Kalman filter for vehicle tracking and a lens rectifica- tion model to correct barrel and pincushion distortion in outdoor cameras.
RadiusAI, Inc.
Feb 202011 mosDec 2020

Jul 20197 mosJan 2020

Navi Mumbai

Developed an understanding of the telecommunication infrastructure and actively contributed to enhancing functionalities at the network and transport layers of the telecom stack.
GitHub
@ojasvita25

Ojasvita’s AI Stack

antigravity
antigravity
elevenlabs
elevenlabs
gemini
gemini
hugging face
hugging face
Qwen
Qwen
RunPod
RunPod

Education

Indian Institute of Technology, Madras

Dual Degree, Electrical and Electronics Engineering

Indian Institute of Technology, Madras

20145 yrs2019

Join Ojasvita and 10k+ other professionals on

Tal Boss

Tal is the app where bosses meet engineers & builders directly, without middlemen

Download Tal Boss

Ojasvita Walgad on Tal

Headline

Machine Learning Engineer | AI x storytelling ❤️

About

Passionate ML/Deep Learning Engineer and Creative Technologist with a strong background in computer vision, LLMs, and multimodal AI. Skilled in Python, PyTorch, and various ML libraries for building, optimizing, and deploying cutting-edge AI models. Experienced in digital world-building by blending advanced technical AI workflows with cinematic visual storytelling across diverse media genres. Github: https://github.com/ojasvita25 Digital Portfolio: https://ojasvita25.github.io/ojasvita/

Location

Bengaluru

AI stack

antigravity

elevenlabs

gemini

hugging face

Qwen

RunPod

Work experience

Creative Technologist at Param Foundation, Sep 2025 - Dec 2025 · 4 mos, On-site, An interactive hide-and-seek installation where visitors playfully test ways to outsmart and evade real-time AI detection. Built using Pygame, YOLOv8 ONNX model, and physical sensor systems to spark curiosity around machine perception and pattern recognition.

AI Futures Fellow at Jaaga, Mar 2025 - May 2025 · 3 mos, Bengaluru South, Karnataka, India, I explored the intersection of AI and storytelling—creating a short film using tools like Suno.ai, ElevenLabs, Eden.art, and Gooey.AI. I created Jengai, a short film that was selected for screening at the Dhurva Film Festival, Darjeeling. My work focused on future-positive narratives, training LoRA models, prompt design, and advocating for the ethical use of AI while addressing inherent biases.

Gen AI Engineer at hedwigAI, Dec 2023 - Apr 2024 · 5 mos, Bengaluru, Karnataka, India, 1. Developed a multi-GPU LoRA fine-tuning framework for a Texify vision encoder-decoder, adapting the model to handwritten math equations to achieve a 46.7% reduction in character edit distance and more than double BLEU score (0.21 →0.44) over baseline. 2. Curated a 10k-sample diverse handwritten dataset combining multi-line formulas and algorithms, designing synthetic noise pipelines (elastic distortion, motion blur) and a contrast-enhancement inference step to resolve degradation in real-world phone captures.

Senior Deep Learning Engineer at RadiusAI, Inc., Dec 2020 - Oct 2023 · 2 yrs 11 mos, Bengaluru, Karnataka, India, Building AI systems for retail stores utilizing advanced computer vision techniques for analyzing customer behavior and managing inventory, successfully deploying them across 700 convenience stores in the US. Experienced with TensorFlow, PyTorch, and deployment frame- works like TensorRT for optimizing and accelerating model infer- ence on NVIDIA GPUs. Proficient in Docker for containerization and Kubernetes (k8s) for container orchestration. Few Projects: 1. Person Re-identification for Tracking in Indoor Cameras - Trained and deployed a Body Part-Based feature learning model for robust people tracking in retail settings. - Achieved significant mAP score improvements: +30.88 points on clients multi-camera data and +12.46 points on public Re-Id dataset (Market 1501), surpassing the Omni-Scale Network (OSNet) re-id model. 2. Multi-Task Network: Monocular Depth and Segmentation - Trained and deployed a multi-task network using a lightweight RefineNet architecture, introducing an enhanced loss function for depth (weighted normal loss and trimmed MAE) and segmentation tasks (weighted Lovasz loss and focal loss). - This model is being used for multiple use-cases: AOI heatmaps (analyzing area-of-interest in retail stores) and 3d interactions (via Voxel Point Clouds). 3. Kalman Tracking and Lens Rectification for Outdoor Cameras - Developed a Kalman filter for vehicle tracking and a lens rectifica- tion model to correct barrel and pincushion distortion in outdoor cameras.

Deep Learning Engineer at RadiusAI, Inc., Feb 2020 - Dec 2020 · 11 mos

Software Developer R&D at Reliance Jio Infocomm Limited, Jul 2019 - Jan 2020 · 7 mos, Navi Mumbai, Developed an understanding of the telecommunication infrastructure and actively contributed to enhancing functionalities at the network and transport layers of the telecom stack.

Education

Dual Degree, Electrical and Electronics Engineering, Indian Institute of Technology, Madras

Public links

github: https://github.com/ojasvita25