Professional Summary
Senior Artificial Intelligence Engineer with master's degree in AI and 6+ years experience in building cutting-edge Vision AI, Gen AI & language solutions across diverse industries – Finance, IT, Agritech, Construction & Automobile. Proficient in architecting, developing and deploying end-to-end ML solutions. Possesses technical expertise & problem-solving skills to tackle challenging AI problems.
Areas of Interest: Machine Learning, Deep Learning, Computer Vision, NLP, MLOps, LLM, BigData.
Work Experience
🤖 Agentic RAG System for Technical Documentation
- Built an Agentic RAG application for QnA with technical documents and engineering drawings
- Researched, evaluated, and fine-tuned multiple language models on industry-specific datasets to enhance specialized knowledge and domain relevance
- Engineered a scalable microservices architecture by containerizing partition, ingestion, and query pipelines with Docker, enabling modular deployment and maintenance
- Deployed and managed the application on Kubernetes, achieving high availability, efficient scaling, and streamlined orchestration of all service components
LLM
VisionLM
Agentic AI
LangChain
LangGraph
MLOps
Azure
🎥 AI Video Analytics for Worksite Safety
- Built AI video analytics solution for safety & productivity usecases at worksite. Implement state-of-the-art object detection, tracking, human and facial pose models, to identify, process and flag alerts with 90+% recall & precision
- Design & implement image-text-to-text Multimodal Vision LLM to identify & flag safety violations for the non-logic based usecases - by analyzing images against preset user configurable textual inputs
- Perform model optimization and quantization to ONNX, TensorRT, NCNN to achieve faster inference
- Create end to end ML pipeline using ClearML to maintain and streamline dataset versioning, training, evalution and benchmarking of the models
PyTorch
OpenCV
ML Pipeline
LLM
Tracking
Pose
MLOps
🍎 SmartGrader - AI Fruit Counting & Grading System
- Design & implement AI for the counting & grading of fruits using image, processed on the edge
- Build object detection, tracking, segmentation, classification model across different agritech solutions, with counting acc. of 99%+, grading acc. of 95%+, significantly improving quality, reduce manpower
- Optimized inference time by quantizing model to reduce size by 40% & inference speed by 60%
- Develop CI/CD pipeline utilizing Docker with Azure Container Registry and Github Actions
Object Detection
Tracking
Segmentation
Classification
TensorFlow
Azure
Docker
CI/CD
💬 Chat with Fruit Doctor - Agricultural LLM Platform
- Implemented a LLM based chat-platform application for the farmers and stakeholders
- Fine-tuned 7-billion parameter LLM model to domain specific dataset to achieve highly accurate model
- Built conversational AI system to provide agricultural advice and plant disease diagnosis
Large Language Model
Fine-tuning
PEFT
LangChain
HuggingFace
PyTorch
🛰️ Super Resolution of Satellite Images
- Designed patented state-of-the-art Super-Resolution GAN model to generate high-resolution images from low-resolution satellite images
- Enhanced satellite imagery for precision agriculture and crop monitoring applications
- Developed innovative GAN architecture specifically optimized for agricultural remote sensing data
GAN
Super Resolution
Remote Sensing
Computer Vision
PyTorch
Patent
🚗 Autonomous Vehicle Vision System
- Built a deep learning model to detect street objects in an image, occluded by raindrops, to flag self-driving mode of an autonomous vehicle on or off, to ensure safety
- Used GANs for image de-raining & applied OpenCV image processing techniques to minimize raindrop effect
- Implemented Siamese neural network to one-shot verify accuracy of predicted objects at inference
- Developed safety-critical computer vision system for autonomous driving in adverse weather conditions
Computer Vision
GAN
Siamese Network
OpenCV
PyTorch
Autonomous Vehicles
🎤 Topic Transition System for Meeting Analysis
- Built a topic-transition & sentiment detection system to study conversation in a meeting-room environment
- Built state-of-the-art Wav2Vec2 based Automatic Speech Recognizer (ASR) dedicated towards Singaporean English accent
- Developed NLP pipeline for real-time conversation analysis and sentiment tracking
Speech Processing
Wav2Vec2
ASR
NLP
Sentiment Analysis
HuggingFace
📄 Financial Document Processing System
- Built image processing pipelines for document image correction using OpenCV & traditional computer vision techniques
- Designed deep learning models for identifying key features within documents; CNNs such as Faster-RCNN, Mask-RCNN
- Enhanced algorithms for data extraction from financial documents; Applied NLP techniques to compare extracted data with ground truth
- Developed Rest API & deployed scalable dockerized Flask applications on AWS
- Independent project handling & direct interaction with multinational bank clients
OpenCV
Faster-RCNN
Mask-RCNN
Document Processing
NLP
Flask
Docker
AWS
⚡ AI-Powered Infrastructure Prediction Engine
- Designed, modelled and maintained an AI-powered prediction engine to prevent and auto-resolve problems like availability, performance, security, capacity across infrastructure stack
- Built, automated and managed machine learning pipeline with HPE Ezmeral to perform data ingestion, cleaning, feature engineering, model training and deployment tasks
- Proposed new data collection points & developed interface to fetch sensor data
- Implemented predictive analytics for proactive infrastructure management and issue resolution
Machine Learning
Predictive Analytics
HPE Ezmeral
MLOps
Infrastructure Monitoring
Data Engineering
Education
Master in Computer Science - Artificial Intelligence specialization
National University of Singapore 2021 - 2022CAP: 4.45/5
Bachelor of Technology - Electrical & Computer engineering
Manipal University 2014 - 2018CGPA: 9.36/10
Technical Skills
Programming Languages
Python
C++
JavaScript
HTML
CSS
AI/ML Libraries
PyTorch
TensorFlow
Keras
OpenCV
LangChain
LangGraph
Transformers
Tools & Frameworks
Docker
Kubernetes
ClearML
MLFlow
Flask
Git
GitHub
Databases & Cloud
MySQL
PostgreSQL
MongoDB
AWS
Azure
Tableau
Spark
Projects
Video Visual Relationship - Video Captioning
Designed a seq2seq architecture to identify object-relationship pairs in a video with top-5 accuracy of 75%. Extracted and sequenced features from video frames using InceptionV4 and ResNet128 CNN models. Modelled bi-directional multi-stacked RNN decoder to unroll caption.
CNN
RNN
InceptionV4
ResNet128
Question-Answering on Stanford SQUAD v2.0
Built a transformer based system to answer questions from Wikipedia reading comprehensions. Explored architectures like BERT, ALBERT, ELECTRA, RoBERTa, XLNet as base model for QA system. Achieved top 35 rank on world leaderboard with F1 score of 87.18 using a custom finetuned RoBERTA-large.
BERT
RoBERTa
Transformers
NLP