Resume

Work Experience

Software Engineer – AI/ML, Sony India Software Centre, Bengaluru, Karnataka, Aug 2024 – Present

Partitioned DeepSeek and Phi-3 models by splitting transformer layers across memory-constrained Raspberry Pi and Jetson Nano for real-time inference; preserved 100% accuracy without compromising architectural integrity.
Deployed an AI-based image failure detection on Jetson Orin NX to detect defocus, off-centre framing, and exposure issues; achieved 82% accuracy, cut drone reshoots by 40%, and was adopted by 3+ U.S. energy clients.

AI/ML Trainee, Sony Finishing School @ IITM Pravartak, Chennai, Tamil Nadu, Feb 2024 – Aug 2024

Led cross-functional development of a CNN-based murmur detection system using Mel-spectrograms from digital stethoscope data, achieving 86% accuracy in early identification of pathological heart sounds for remote diagnosis.
Built a quality-gating pipeline for scanned legal documents using Otsu’s thresholding and custom blur detection; flagged low-quality pages for enhancement, reducing OCR extraction errors by 20% across 15K legal documents.
Secured a full-time offer from Sony India post-program, as a top 1% candidate based on performance and interview.

Associate Software Engineer – I, HighRadius, Hyderabad, Telangana, Jul 2022 – May 2023

Developed a LightGBM-based payment date prediction model using customer behavioral trends and historical transactions; enabled proactive prioritization of delinquent accounts, improving overdue invoice recovery by 25%.
Engineered a CatBoost-based classification system to evaluate deduction validity across 600k+ financial transactions; isolated high-confidence invalid claims with 93% accuracy, eliminating 73% of manual follow-ups.

Product Intern – Data Science, HighRadius, Bhubaneswar, Odisha, Aug 2021 – Jun 2022

Eliminated SQL bottlenecks in a high-throughput, asynchronous Java REST API powering ML-based payment prediction pipelines; reduced latency by 40% using CompletableFuture, batched queries, and MySQL indexes.
Crafted a scalable feature engineering pipeline leveraging Python (Pandas) and SQL to efficiently extract, transform, and aggregate critical variables from large transactional data, boosting ML model accuracy by 8%.

Designed a domain-specific RAG chatbot using LangChain, Ollama, and LLaMA 3.2 (7B) to generate semantically grounded responses, achieving 87% answer accuracy on a benchmark of 100 curated health and fitness queries.
Optimized chunking and FAISS index to reach 92% top-3 document retrieval accuracy on 50+ medical PDFs.
Configured cloud-independent LLaMA 3.2 inference, achieving sub-2s latency on constrained healthcare systems.

Constructed a YoloV1 object detection model optimized for real-time applications using PyTorch.
Streamlined dataset handling and training workflows, enhancing model training and evaluation efficiency.
Integrated MLflow for experiment tracking, metrics logging, and model performance visualization.

Programming Languages & Tools: Python, C/C++, SQL, Git, Docker, Kubernetes, AWS (SageMaker, S3)
Data Engineering & Processing: Pandas, NumPy, PySpark, MongoDB, Apache Kafka, Airflow
Machine Learning: Scikit-learn, TensorFlow, PyTorch, ONNX, MLflow, FastAPI, Hugging Face Transformers
Large Language Models: OpenAI API, Anthropic Claude, LangChain, LlamaIndex, RAG, llama.cpp, FAISS, Chroma

B.Tech in Electronics and Communications Engineering (Hons.), Lovely Professional University, Punjab, India
Graduation: June 2022

CGPA: 8.44
Relevant Coursework: Introduction to Artificial Intelligence, Digital Signal Processing, Database Management System, Workshop on Image Processing, Computer Programming, Data Structures and Algorithms