Aniruddha Kalkar

Senior Machine Learning Engineer | AI Enthusiast | Software Developer

About Me

I am a seasoned Machine Learning Engineer with over 5 years of experience in developing AI-driven solutions that tackle real-world challenges. My journey spans advanced research, cutting-edge AI deployments, and scalable system architectures. Currently leading machine learning efforts at Opal AI Inc., I specialize in crafting multimodal models, designing scalable backends, and delivering systems capable of processing millions of data points with precision and efficiency.

Throughout my career, I’ve had the privilege of working on transformative projects like improving object detection accuracy for the US Department of Transportation, enhancing 3D LiDAR segmentation models, and building high-performance pipelines that automate workflows and save thousands of hours annually. My expertise lies in bridging the gap between research and implementation, ensuring that innovative ideas translate into impactful, real-world solutions.

Experience

Senior Machine Learning Engineer

Opal AI Inc (Aug 2023 - Present)

  • Designed backend for video intelligence platform using Multimodal LLMs on GCP, reducing report generation time by 95%.
  • Built scalable architecture with microservices, databases, and APIs to process 100,000+ minutes of video monthly.
  • Led ML efforts for US Department of Transportation, achieving 92%+ object detection precision for 60+ road elements.

Machine Learning Engineering Intern

Blackberry Corporation (Oct 2022 - Apr 2023)

  • Automated ML workflows using Apache Airflow, saving 2190 hours/year.
  • Optimized data pipelines, reducing latency by 3.61 seconds per batch for 10M files daily.
  • Contributed to scalable ML model deployment and improved runtime performance.

Software Engineering Specialist

Dassault Systèmes Solutions Lab (June 2019 - July 2021)

  • Redesigned UI for CI/CD pipeline, increasing adoption by 63%.
  • Developed REST APIs with 3x faster response times, reducing execution time.
  • Contributed to scalable CI/CD infrastructure, reducing build queue times by 30%.

Research Experience

Researcher

Locomotor Control Lab @ USC (Jan 2022 - Apr 2023)

  • Enhanced VR game to improve skilled locomotion for individuals with neurological impairments.
  • Introduced functionality to store additional user action data for deeper analysis.

Researcher

ICAROS @ USC (May 2022 - Dec 2022)

  • Designed and executed experiments for training 8 Quality Diversity Algorithms in various reinforcement learning environments.
  • Engineered a high-performance Python script for distributed computing with 100 CPUs.

Research Intern

Tata Consultancy Services Research and Innovation (Dec 2018 - Apr 2019)

  • Collaborated on 3 Computer Vision projects, co-authoring a WACV 2020 publication.
  • Introduced optical flow for AR applications, achieving up to 50x Temporal Coherence improvement in label placement.

Publications

Projects

Multi-Teacher Knowledge Distillation for VQA

Capstone project: Reduced model size by 65x and achieved up to 8x inference speed improvement.

  • Tools: Python, PyTorch, OpenCV
  • Optimized knowledge distillation for Visual Question Answering Systems.
UI Template Image to Code Generation

Developed a web app to generate HTML code from UI screenshots.

  • Tools: Python, Flask, Keras, OpenCV, HTML, CSS3
  • Reduced code writing time by 23.4 mins per web page on average.
Face Sketch to Photo-Realistic Image Generation

Created a system to convert hand-drawn sketches into photo-realistic images.

  • Tools: Python, Flask, Keras, OpenCV, HTML, CSS
  • Achieved 77.65% similarity with original images and 87.38% age group prediction accuracy.
Toxic Comment Classification

Classified social media comments into 6 levels of toxicity using NLP.

  • Tools: Python, Keras, Scikit-Learn, NLTK
  • Implemented RNNs to classify word embeddings from social media comments.
Driver Distraction Detection

Developed a tool to detect driver distractions using video feeds and deep learning.

  • Tools: Python, TfLearn, OpenCV, Flask
  • Achieved 91.08% accuracy in recognizing 10 pre-determined distractions.
Quora Question Pair Similarity

Predicted the similarity between Quora questions using a dual LSTM model.

  • Tools: Python, Keras, Scikit-Learn, NLTK
  • Used dual LSTM heads to generate word embeddings and calculate similarity scores.

Skills

Programming Icon

Programming Languages

Python, JavaScript, C/C++, Java, C#, HTML, CSS, React, Angular.js, Node.js, React.js, GoLang

Programming Frameworks

Flask, Django, Unity, RESTful APIs, gRPC

Machine Learning Icon

Machine Learning, AI & NLP Frameworks

TensorFlow, PyTorch, Keras, OpenCV, matplotlib

Database Icon

Databases

Firestore, SQL, MySQL, MongoDB, DynamoDB, AWS S3

Cloud Tools Icon

Cloud Tools & Platforms

Vertex AI, GCP Cloud Run, GCP Cloud Functions, GCP API Gateway

Google Cloud Platform, AWS EC2, AWS Lambda

Model Deployment Icon

AI Model Deployment Tools

AWS SageMaker, AirFlow, MLFlow, Prefect

CI/CD Icon

CI/CD Tools

Docker, Kubernetes, Kubeflow, Metaflow

GitHub Actions, Terraform, Jenkins, Bitbucket

Contact

Email Phone GitHub LinkedIn Resume