Summary
Machine Learning Engineer with extensive experience in Computer Vision. Spearheaded a team of up to 14 Computer Vision and Data Engineers during a dynamic 3x growth phase. Collaboratively developed and successfully delivered the initial iteration of a cutting-edge product offering, encompassing more than 90 real-time Video Analytics solutions, seamlessly integrated across a global network of 20K+ CCTV cameras.
Tech Stack
With 3+ years of industry experience building scalable Computer Vision Products and 2+ years of academic research experience leveraging the latest advancements in the field, I have honed my skills in computer vision, machine learning, and cutting-edge research methodologies. Let's explore how my expertise can contribute to your projects and drive innovation.
Testimonials
Explore feedback from mentors, managers, professors, and colleagues regarding my abilities as a Machine Learning Engineer and Lead. These testimonials underscore my dedication, expertise, and professionalism, positioning me as an ideal candidate for any organization. They serve as proof of my readiness to excel in a corporate environment.
Projects
I learn by building! Allow me to introduce you to some of my fun projects, crafted over the years, often during my weekends. :-)
Autonomous Driving
End-to-end Conditional Imitation Learning in a Real-World model city.
Resume
Research Experience
Graduate Research Assistant
Jan 2023 - Present
H2X Lab, Boston University, Boston, MA
- Develop zero-shot Sim2Real using foundation models like SegmentAnything and DINOv2 to directly translate learned controls from CARLA simulator to the real world.
- Applied test-time dropout to Transfuser (Chitta et al.) pre-trained models to modify model architecture and performance, and to examine the correlation between online and offline evaluation metrics for 36 routes spanning 6 towns in the CARLA simulator.
- Experimented with sensor fusion using vision and LIDAR-based multi-modal conditional imitation learning incorporating auxiliary tasks such as depth estimation and semantic segmentation for autonomous driving in CARLA simulator.
- Explored RegNet and SampleRNN for audio generation from visual scenes for representation pre-training of navigation agents.
Graduate Research Assistant
Feb 2023 - May 2023
BIT Lab, Boston University, Boston, MA
- Developed rule-based multi-modal algorithm that leverages text prompts, image tags, and visual features to assist causal inference on user art study, enabling deeper analysis of user behavior and preferences.
- Developed ViT and DINOv2-based models using PyTorch to identify AI-generated Deviant Art and achieved an accuracy of 92.04%.
Undergraduate Research Assistant
Feb 2018 - Jun 2019
RNS Institute of Technology, Bangalore, India
- Authored 4 research papers with 100+ citations; performed comparative study in preprocessing techniques and algorithmic survey in sentiment analysis, forecasting, and encoding.
Education
MS - Artificial Intelligence
2022 - 2024
Boston University, Boston, MA, USA
Research Assistant: H2X Lab and BIT Lab
Courses: Robot Learning and Vision for Navigation, Computer Vision, Geometric Processing, Data Science Tools and Applications, Principles of Machine Learning, Artificial Intelligence.
BE - Electronics Engineering
2015 - 2019
Visvesvaraya Technological University, Bangalore, India
Project: Automatic Helmetless Rider Detection using Deep Learning
- "Best Outgoing Student - 2019" among 180+ students
- "First prize" in state competition at IIIT-Bangalore
- "Letter of Appreciation" from the HoD, dept. of ECE
Professional Experience
Machine Learning Engineer (Contractor)
Jun 2023 - Aug 2023
PRADCO - Outdoor Brands, Remote, USA
- Experimented with and built algorithms for detection, segmentation, genearative AI, and 3D computer vision.
- Confidential/sensitive information withdrawn.
Computer Vision Engineer & Lead
Jun 2019 - Jun 2022
Wobot Intelligence (Wobot.ai), New Delhi, India
- Spearheaded a team of 14 engineers to develop over 90 real-time video analytics solutions scaled on Cloud using Kubernetes for 200+ concurrent CCTV cameras, resulting in increased hygiene compliance by 2x in the food and hospitality industry.
- Enforced safety & hygiene compliance by developing multi-object detection & tracking, pose estimation, activity recognition, person re-identification, and face recognition algorithms, deployed across 3 continents reducing non-compliance by 25%+.
- Applied classification, object detection & tracking algorithms like ResNet, Inception, EfficientNet, EfficientDet, YOLO, Centroid Tracking, and OpenCV Tracking to satisfy product requirements based on available compute resources.
- Reduced data-to-production time by building development tools for data and models (using Python, Tensorflow, PyTorch & OpenCV) resulting in a 3x increase in productivity, positively impacting the team's efficiency and reducing time-to-market by 50%.
- Implemented Synthetic Dataset Generation for object detection, reducing labeled data requirements by 35% and accelerating computer vision model development, resulting in significant cost savings and faster time-to-market.
- Improved alert precision by up to 95% using ensemble models and temporal features reducing false positive alerts by 30%.
Contact
Thank you for visiting my website! I'm excited to hear from you. Whether you have questions, want to collaborate, or simply want to say hello, feel free to reach out to me through the email below.
Location
Boston, MA 02134
animikh@bu.edu
Phone
+1 (857) 260-0017