Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I’m Manish Kumar, an AI Engineer with 5.5+ years of experience building generative AI systems, especially text-to-video pipelines using diffusion transformers and spatio-temporal modeling. I enjoy turning ideas from vision and language into practical, high-performance tools—whether that’s creating large synthetic video datasets for robotics training, or designing systems that learn to predict and plan in complex environments. In my recent work at Alphadroid, I’ve led end-to-end development of Sim2Real video generation and world-model-driven robotic planning, while also optimizing inference for low latency on edge devices. Previously at Oodles Technologies, I helped lead the transition to Generative AI for industrial predictive maintenance, built real-time computer vision pipelines, and scaled multi-sensor data platforms—always with an eye on robustness, speed, and real-world impact.…I’m Manish Kumar, an AI Engineer with 5.5+ years of experience building generative AI systems, especially text-to-video pipelines using diffusion transformers and spatio-temporal modeling. I enjoy turning ideas from vision and language into practical, high-performance tools—whether that’s creating large synthetic video datasets for robotics training, or designing systems that learn to predict and plan in complex environments. In my recent work at Alphadroid, I’ve led end-to-end development of Sim2Real video generation and world-model-driven robotic planning, while also optimizing inference for low latency on edge devices. Previously at Oodles Technologies, I helped lead the transition to Generative AI for industrial predictive maintenance, built real-time computer vision pipelines, and scaled multi-sensor data platforms—always with an eye on robustness, speed, and real-world impact.

Manish Kumar

AI Engineer, Web Developer, Software Engineer, +2





I’m Manish Kumar, an AI Engineer with 5.5+ years of experience building generative AI systems, especially text-to-video pipelines using diffusion transformers and spatio-temporal modeling. I enjoy turning ideas from vision and language into practical, high-performance tools—whether that’s creating large synthetic video datasets for robotics training, or designing systems that learn to predict and plan in complex environments. In my recent work at Alphadroid, I’ve led end-to-end development of Sim2Real video generation and world-model-driven robotic planning, while also optimizing inference for low latency on edge devices. Previously at Oodles Technologies, I helped lead the transition to Generative AI for industrial predictive maintenance, built real-time computer vision pipelines, and scaled multi-sensor data platforms—always with an eye on robustness, speed, and real-world impact.…I’m Manish Kumar, an AI Engineer with 5.5+ years of experience building generative AI systems, especially text-to-video pipelines using diffusion transformers and spatio-temporal modeling. I enjoy turning ideas from vision and language into practical, high-performance tools—whether that’s creating large synthetic video datasets for robotics training, or designing systems that learn to predict and plan in complex environments. In my recent work at Alphadroid, I’ve led end-to-end development of Sim2Real video generation and world-model-driven robotic planning, while also optimizing inference for low latency on edge devices. Previously at Oodles Technologies, I helped lead the transition to Generative AI for industrial predictive maintenance, built real-time computer vision pipelines, and scaled multi-sensor data platforms—always with an eye on robustness, speed, and real-world impact.

Available to hire

I’m Manish Kumar, an AI Engineer with 5.5+ years of experience building generative AI systems, especially text-to-video pipelines using diffusion transformers and spatio-temporal modeling. I enjoy turning ideas from vision and language into practical, high-performance tools—whether that’s creating large synthetic video datasets for robotics training, or designing systems that learn to predict and plan in complex environments.

In my recent work at Alphadroid, I’ve led end-to-end development of Sim2Real video generation and world-model-driven robotic planning, while also optimizing inference for low latency on edge devices. Previously at Oodles Technologies, I helped lead the transition to Generative AI for industrial predictive maintenance, built real-time computer vision pipelines, and scaled multi-sensor data platforms—always with an eye on robustness, speed, and real-world impact.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Language

English

Fluent

Hindi

Intermediate

Work Experience

Tech AI Lead at Alphadroid, Noida, Uttar Pradesh

May 1, 2024 - Present

Led the design of video generation pipelines using diffusion transformers and spatio-temporal attention, producing 500K+ diverse training clips that improved robotic manipulation robustness by 67%. Directed Sim2Real training frameworks that combine generative video synthesis with simulation rollouts, reducing physical data collection needs by 70%. Integrated world model architectures (Dreamer, COSMOS-style) with NeRF and 3D Gaussian Splatting into navigation and perception modules for predictive planning and context-aware behavior. Optimized inference using CUDA kernel fusion, INT8 quantization, and pruning, reducing edge-device latency from 800ms to 120ms.

Tech AI Lead at Alphadroid

May 1, 2024 - Present

Led the design and implementation of video generation pipelines using diffusion transformers and spatio-temporal attention, generating 500K+ diverse training clips that improved robotic manipulation robustness by 67%. Directed Sim2Real training frameworks by combining generative video synthesis with simulation rollouts, reducing physical data collection needs by 70%. Integrated world model architectures (Dreamer, COSMOS-style) with NeRF and 3D Gaussian Splatting into navigation and perception modules to enable predictive planning and context-aware behavior. Optimized video inference for edge deployment using CUDA kernel fusion, INT8 quantization, and pruning, reducing latency from 800ms to 120ms.

Senior AI Development at Oodles Technologies, Gurugram, Haryana

January 1, 2023 - May 1, 2024

Led the shift from traditional ML to Generative AI by integrating GPT-4 and Claude into a Wellsite platform, cutting equipment failure prediction errors by 35% and helping prevent over $2M in potential losses. Designed and implemented computer vision pipelines with OpenCV, YOLOv8, and custom CNN models for real-time industrial monitoring, achieving 96% object detection accuracy across varying lighting conditions. Built scalable data processing systems using Apache Kafka and Druid for 20GB+ multi-sensor data daily, and mentored five junior developers on MLOps and distributed systems best practices. Drove technical decisions for multiple cross-functional AI projects by setting coding standards and review processes, improving delivery speed by 40% and reducing production bugs by 50%.

Senior AI Development at Oodles Technologies

January 1, 2023 - May 1, 2024

Drove a shift from traditional ML to Generative AI by integrating GPT-4 and Claude into the Wellsite platform, reducing equipment failure prediction errors by 35% and preventing $2M+ in potential losses. Designed and implemented real-time computer vision pipelines using OpenCV and YOLOv8, achieving 96% object detection accuracy across challenging lighting and environmental conditions. Built scalable data processing systems with Apache Kafka and Druid to handle 20GB+ multi-sensor data daily, and mentored five junior developers on MLOps and distributed system design. Set technical standards and led reviews for multiple AI projects, improving delivery speed by 40% and reducing production bugs by 50%.

Associate Development at Oodles Technologies, Gurugram, Haryana

March 1, 2021 - January 1, 2023

Developed a document retrieval system using transfer learning with PaLM and YOLO object detection, achieving 93% classification accuracy and automating data extraction. Automated invoice processing workflows for logistics clients using intelligent OCR and NLP, reducing manual processing time from 8 hours to 2 hours per batch. Built and improved production-grade AI data pipelines and extraction logic to support end-to-end automation.

Associate Development at Oodles Technologies

March 1, 2021 - January 1, 2023

Built a document retrieval system using transfer learning with PaLM and YOLO object detection, achieving 93% classification accuracy and fully automating data extraction. Developed invoice processing automation for logistics clients, reducing manual processing time from 8 hours to 2 hours per batch through intelligent OCR and NLP workflows.