5+ years at Samsung R&D turning research into things people actually use — 7 patents co-authored, peer-reviewed work published at IEEE GLOBECOM '21.
Corporate Career
Mar '24 – Present
Software Engineer II (Lead Engineer)
Stellar Project Award · Q1 '26
Challenge Project Award · '24
- Currently leading the architectural design of backend systems and core development of ML-driven image processing models for Automated Defect Analysis and Conversational AI chatbot.
- Engineered a cost-optimized AR Assisted TV Installation System (PWA) for Service Engineers, replacing manual estimations with precise drill-hole mapping and customer previews.
- Scaled the system to support nation-wide installations, utilizing on-device image processing to eliminate heavy WebRTC streams & CUDA-accelerated overheads to reduce infra costs.
- Overall reduction in installation time by 33% and Won the Stellar Project Award (Q1 '26) in Best Growth (Innovation) Category.
- Architected a multi-cloud Generative AI Gateway Server on the Samsung Cloud Platform (SCP), serving as a centralized LLM integration hub for AI-powered Summaries and Responses in the Email Management System.
- Implemented cost-aware dynamic traffic routing between AWS & Azure OpenAI; GenAI features decreased Agent Average Handling Time (AHT) by 25% and drove an INR ~7.5M/year OpEx reduction; earned the Challenge Project Award '24.
- Evaluated a Generative AI-based call transcription and evaluation POC; advised against production as LLM inference and transcription costs for 2,700+ daily audio hours negated efficiency gains.
- Designed the Machine Learning based pipelines and system architectures for 4 co-authored patents across Extended Reality (XR), AI-generated Video Processing, and Machine Learning.
Aug '21 – Feb '24
Software Engineer I
- Architected the Total Solution Center, a unified platform consolidating fragmented CRM data and omni-channel interactions (Voice, Email, WhatsApp), cutting Agent AHT by 30%.
- Maintained the Image Analytics System against OWASP and package-level CVEs and upgraded image manipulation detection to MVSS-Net++, tampered image submissions reduction by 85%.
- Prototyped a Federated Learning POC for Samsung Keyboards on TV, to drive external research partnerships. Evaluated internal technical proposals on the SRI-D AI Review Board.
- Formulated the core ML logic and technical architectures for 3 co-authored patents across Machine Learning and Dialogue Systems.
Research & Internships
Jan '21 – Jun '21
Spring Research Intern
with Jigar Halani & Sunil Patel
- Accelerated Deep Learning deployment pipelines by quantizing PyTorch models (SSD, ResNet-50) to INT8 precision and converting them to ONNX, minimizing latency and memory footprint.
- Deployed optimized models at scale in production utilizing NVIDIA TensorRT and Triton Inference Server.
- Built end-to-end AI-powered video analysis applications using the DeepStream SDK and GStreamer pipelines (C++/Python).
Jul '20 – May '21
Undergraduate Researcher · Bachelor's Thesis
- Proposed Borrowing Energy with Adaptive Rewards (BEAR), a novel Deep Reinforcement Learning (Actor-Critic) algorithm to maximize throughput in borrowing-aided Energy Harvesting (EH) systems.
- Formulated an adaptive reward and penalty function based on parameterized Gaussian distributions to optimize energy scheduling and prevent energy hoarding.
- Achieved a 35.45% enhancement in sum throughput over baseline non-borrowing models; published and presented findings at IEEE GLOBECOM '21.
Apr '20 – Jun '20
Summer Research Intern
Speech and Language Technology Lab, Saarthi.ai
with Vishwanath Jha & Akash Singh
- Engineered a Deep Reinforcement Learning pipeline incorporating the Actor-Critic framework and hybrid value networks to optimize task-oriented dialogue policies.
- Implemented distributed RL for parallel trajectory sampling on SGDS-DSTC datasets, scaling a pre-training architecture for multi-agent policy models.
May '19 – Jul '19
Research Intern
Pattern Recognition and Image Analysis Lab, IIT Indore
with Dr. Vivek Kanhangad
- Developed a customized pore-based fingerprint recognition system using a multitask deep convolutional neural network (ResNet/Inception) to extract fixed-length representations from high-resolution images.
- Formulated an Unsupervised Deep Domain Adaptation methodology utilizing backpropagation to enable robust cross-sensor pore detection.
- Evaluated network performance by calculating cosine similarity scores between representations to generate probability distributions of genuine and imposter matches.
College Activities
Cyberwarp Module Head
Apr '21
Tecnoesis '21, NIT Silchar
Board Member
Oct '20 – May '21
Obiettivo, Photography Club, NIT Silchar
Technical Head · Machine Learning
Sep '19 – Jul '20
Robotics Club, NIT Silchar
Event Head · Data Strata
Feb '20
Tecnoesis '20, NIT Silchar
Publicity Co-Head
Dec '19 – Feb '20
Tecnoesis '20, NIT Silchar
Web Dev Head
Oct '19 – Nov '19
SPIC MACAY, NIT Silchar
Technical Head
Sep '19
All India Inter NIT Tournament
Winner · Hackathon Module
Aug '19
NIT Conclave '19, NIT Rourkela
Organized amongst all 31 National Institutes of Technology (NITs)
National Level Finalist
Mar '19
Smart India Hackathon '19, NIT Warangal
Organized by MHRD, Government of India