About

Education

IIT Kharagpur

M.Tech in Computer Science and Data Processing

July 2021 | Kharagpur, WB

CGPA: 9.11 / 10.0

Work Experience

Senior AI Engineer – Justdial

Jan 2024 – Present · Bengaluru, India (On-site)

• Optimized inference for large language and vision models like Llama3.1 8B, Llava1.5v 7B, and Llama3.2-vision 11B using Nvidia's Triton Inference Server. Integrated TensorRT backend and KV cache to reduce latency by 74.12% (from 17s to 4.4s). Currently implementing Flash Attention and refining autoscaling with Ray Serve.
  Tech Stack: TensorRT-LLM, Nvidia Triton Inference Server, PyTorch, FastAPI, Docker, HuggingFace Transformers
• Designed and deployed a system to dynamically display profile images based on user-searched categories. Used fine-tuned SBERT, MLflow for tracking and deployment, and Milvus for free-text vector search. Improved CTR by 16.37% within a month.
  Tech Stack: SBERT, Milvus, Llama3.2-vision 11B, OpenCV, Guardrails
• Developed an on-premise video transcoding pipeline for HTTP Live Streaming (HLS), reducing costs by avoiding cloud services.
  Tech Stack: RabbitMQ, Docker, FFmpeg, Nvidia NVENC unit
• Led the design and deployment of a centralized MLflow system to support reproducibility, collaboration, and model lifecycle management across teams.
  Tech Stack: MLflow, AWS S3, MySQL, Nginx-Proxy server
• Built an image enhancement pipeline using GAN models (RealESRGAN_x4plus, RealESRNET_x4plus, GFPGAN), integrated with AWS S3. Enhanced over 1M images, improving overall user experience.
  Tech Stack: GANs, AWS S3
• Designed and deployed image and video metadata extraction system, processed over 100M images and 1M+ videos across major Indian cities.
  Tech Stack: Docker, Flask, Gunicorn, MongoDB, MySQL, Apache Kafka, Redis, Go, Python, MediaPipe, ImageMagick

Data Engineer Intern – CropIn

Nov 2023 – Jan 2024 · Bengaluru, India (On-site)

• Built ResNet9 models for plant-disease prediction and treated field photos — not lab datasets — as the real test.
• Diagnosed domain shift: models that scored highly on clean images collapsed on real farm photos because of background variation and class imbalance.
• Closed part of the gap with blended/augmented data, regularization, dropout, and weighted loss, and used deployment accuracy as the metric that mattered.

Projects

• On-premise HLS video transcoding — Queue-driven FFmpeg + NVENC pipeline that transcodes video locally for HTTP Live Streaming, avoiding cloud transcoding cost.
• Cross-team MLflow infrastructure — Centralized tracking and model registry (MLflow, S3, MySQL, Nginx) so teams can reproduce runs and share models.
• Profile / category image search — Serves the most relevant business photos for a search query using CLIP/SBERT embeddings, Milvus, and vision LLMs.
• Image upscaling and restoration — GPU/CPU workers that enhance faces, text, and general photos with GFPGAN, RealESRNet, and RealESRGAN, then write results back to S3.

Skills

Programming Languages

• Expert: C++, Python
• Intermediate: C, SQL

Tools

• Databases & Queues: MySQL, MongoDB, RabbitMQ, Apache-Kafka
• Utilities: Postman, Git, GitHub, GitLab (version control), Jupyter Notebook / JupyterLab
• DevOps & Cloud: Docker, Amazon Web Services (AWS), CI/CD Pipeline

Frameworks and Libraries

• Web Frameworks: Flask, FastAPI
• Data Processing & Visualization: Numpy, Pandas, Matplotlib, Seaborn
• Machine Learning & Deep Learning: Scikit-learn, Keras, TensorFlow, OpenCV, Pytorch
• Natural Language Processing (NLP): NLTK, SpaCy, Transformers (Hugging Face)

Publications

• V. G. Nandanwar, M. Mohammed Kashif, and R. S. Ankushe, “Portable weight measuring instrument,” in Proc. 2017 Int. Conf. Recent Trends in Electrical, Electronics and Computing Technologies (ICRTEECT), 2017, pp. 44–48. DOI: 10.1109/ICRTEECT.2017.23