About Me
ML Platform Engineer with 3+ years building and operating production ML infrastructure and MLOps pipelines. Delivered end-to-end platform capabilities including automated feature engineering, model training orchestration, API-based serving, and monitoring on cloud. Shipped scalable forecasting systems with measurable platform reliability and business impact.
Career Journey
Jan 2024 - Dec 2025
M.S. in Computer Science
Syracuse University
Syracuse, NY
Specialization: Machine Learning, AI, and Reinforcement Learning
Jan 2024 - Dec 2025
M.S. in Computer Science
Syracuse University
Syracuse, NY
Specialization: Machine Learning, AI, and Reinforcement Learning
Jan 2021 - Dec 2023
ML Engineer
Accenture
Bangalore, India
- Designed and built a scalable ML platform for demand forecasting, delivering automated data ingestion, feature engineering pipelines, XGBoost model training, and batch inference infrastructure supporting 50K+ SKU forecasts daily.
- Implemented REST API-based model serving with scheduled ETL workflows on Azure, enabling downstream applications and reporting tools to consume predictions reliably; reduced manual deployment effort through integrated CI/CD for model versioning and safe rollouts.
- Developed monitoring and logging infrastructure to track model performance, data quality, and pipeline health; automated alerts for drift detection and failures, improving platform reliability and reducing incident response time.
- Collaborated with procurement and operations stakeholders to define forecast horizons, evaluation metrics, and business rules; delivered Power BI dashboards for 15+ users, reducing stockouts by 35% and excess inventory costs by $100K annually.
- Architected end-to-end platform components across ingestion, feature management, training orchestration, inference serving, and monitoring to ensure scalability, maintainability, and alignment with enterprise MLOps best practices.
Jan 2021 - Dec 2023
ML Engineer
Accenture
Bangalore, India
- Designed and built a scalable ML platform for demand forecasting, delivering automated data ingestion, feature engineering pipelines, XGBoost model training, and batch inference infrastructure supporting 50K+ SKU forecasts daily.
- Implemented REST API-based model serving with scheduled ETL workflows on Azure, enabling downstream applications and reporting tools to consume predictions reliably; reduced manual deployment effort through integrated CI/CD for model versioning and safe rollouts.
- Developed monitoring and logging infrastructure to track model performance, data quality, and pipeline health; automated alerts for drift detection and failures, improving platform reliability and reducing incident response time.
- Collaborated with procurement and operations stakeholders to define forecast horizons, evaluation metrics, and business rules; delivered Power BI dashboards for 15+ users, reducing stockouts by 35% and excess inventory costs by $100K annually.
- Architected end-to-end platform components across ingestion, feature management, training orchestration, inference serving, and monitoring to ensure scalability, maintainability, and alignment with enterprise MLOps best practices.
Jul 2019 - Dec 2019
Computer Vision Intern
Lenskart
Bangalore, India
- Built a computer vision pipeline for real-time virtual eyewear try-on using facial landmark detection and custom models to align frames with user faces.
- Implemented geometric alignment and frame placement logic to improve fit consistency across varied face shapes.
- Integrated the pipeline into Android using native camera APIs and JNI, reducing preview latency by 35% and improving runtime stability on mobile devices.
Jul 2019 - Dec 2019
Computer Vision Intern
Lenskart
Bangalore, India
- Built a computer vision pipeline for real-time virtual eyewear try-on using facial landmark detection and custom models to align frames with user faces.
- Implemented geometric alignment and frame placement logic to improve fit consistency across varied face shapes.
- Integrated the pipeline into Android using native camera APIs and JNI, reducing preview latency by 35% and improving runtime stability on mobile devices.
Aug 2016 - May 2020
B.Tech in Electronics and Communication Engineering (ECE)
Birla Institute of Technology and Science, Pilani (BITS Pilani)
Aug 2016 - May 2020
B.Tech in Electronics and Communication Engineering (ECE)
Birla Institute of Technology and Science, Pilani (BITS Pilani)
Project Portfolio
Fash-ON | Agentic AI Stylist + Recommendation Platform
GNN
ResNet-50
Docker
GCP
CI/CD
Microservices
LangGraph
- Built a Graph Neural Network recommender with ResNet-50 embeddings and custom graph convolution layers, using multi-task learning to reach 91% accuracy (Polyvore) and 87% (Fashion-Gen).
- Containerized the model inference service with Docker and deployed on Google Cloud Run with automated CI/CD pipelines; implemented token-based authentication, API versioning, and performance monitoring to handle 400+ daily requests with sub-200ms latency.
- Developed a RESTful API microservice architecture for model serving and user session management; integrated with LangGraph-based agent orchestration for vision-language model feedback, delivering structured outfit scoring and recommendations.
Autonomous Underwater Vehicle | MIT AUVC 2025 Competition Winner
Reinforcement Learning
MuJoCo
Unity
Robotics
Simulation
RL
- Developed a hybrid MuJoCo-Unity simulator for training underwater vehicle RL policies, integrating physics-accurate hydrodynamic forces with visual perception and domain randomization for robust sim-to-real transfer.
- Designed multi-component reward functions and leveraged multi-agent adversarial RL with self-play to improve robustness under noise, running controlled ablations to isolate reward terms that improved success rate and tracking stability.
- Integrated the final policy into the on-vehicle control loop with safety checks and classical fallbacks, validating in pool tests and securing a first place finish at MIT AUVC 2025.
Sub-Quadratic Vision Transformers | Research Project
Vision Transformers
CUDA
Attention Mechanisms
Deep Learning
Research
- Implemented RippleAttention and HydraAttention mechanisms from scratch to replace Multi-Head Attention in Vision Transformers, reducing computational complexity from O(n^2) to O(n) while achieving +13% accuracy on CIFAR-100 and +10% on Tiny ImageNet.
- Engineered custom CUDA kernels for linear attention layers, accelerating training convergence by 2x over baseline ViT architectures through optimized memory access patterns and parallelized matrix operations.
SU Buddy
React
Node.js
JavaScript
MongoDB
Vite
- Developed a comprehensive academic platform using React and Vite for the frontend and Node.js/Express backend, implementing user authentication, data management, and responsive UI components.
- Designed and implemented a MongoDB database structure with efficient data models for student information, academic resources, and institutional data, enabling scalable data management.
- Built a secure RESTful API with JWT, password hashing, input validation, and security headers to ensure safe client-server communication.
Technical Skills
Programming & ML Frameworks
Python
SQL
Java
PyTorch
TensorFlow
Scikit-learn
XGBoost
Data & ML Systems
Feature Engineering
ETL Pipelines
Pandas
Spark
Batch & Online Inference
API-based Model Serving
MLOps & Deployment
Docker
Google Cloud Run
Azure ML
Model Versioning
Monitoring
CI/CD with GitHub Actions
Platform & Software Engineering
Scalable ML Services
Microservices
RESTful APIs
SDLC
Agile Development