May 2025 – Dec 2025 · 7 mos · Austin, Texas, United States · On-site

Jan 2025 – May 2025 · 4 mos · Worcester, Massachusetts, United States

• Developing and Grading Assignment for the Course - ML for Robotics
• Assignments Developed:
• 1. Joint Effort - Predicting the pose (Intro to ML and DL using Action space)
• 2. Push & Learn: Learning Dynamics through learning (Physics Informed Learning)
• 3. Act or Critic: Learning to Pick (Reinforcement Learning)
• 4.ACT of Insertion -- Imitation Learning for Peg Placement and Insertion (Imitation Learning)

Jan 2024 – May 2025 · 1 yr 4 mos · Worcester, Massachusetts, United States

• Developing End-End learning based Robot Grasping techniques
• Thesis Link: https://digital.wpi.edu/concern/etds/m039k929x?locale=en
• 1. Developed Monocular Depth Estimation based Grasping Pipelines
• 2. Developed Gaussian Splatting Pipelines for Real time 3D Reconstruction of Objects intended to Grasp
• 3. Developed Computer Vision + LLM (Multimodal) pipelines for Detection of objects for Grasping
• 4. Combined 1,2 and Traditional Grasping using 3 for a Comprehensive Pick and Place Solution @ELPIS Lab

Sep 2023 – Feb 2024 · 5 mos · Worcester, Massachusetts, United States

• 1. Worked on Developing Monocular camera based optical flow navigation for Aerial Vehicles.
• 2. Implemented the Overall Stack of Peering based Navigation on Custom Drone Hardware.
• 3. Conducted Multiple experiments to prove with this approach it should be feasible to avoid Static, Dynamic Obstacles and also pass through Gaps.

Jun 2024 – Aug 2024 · 2 mos · Philadelphia, Pennsylvania, United States · Remote

• 1. Experimented and Evaluated Open Source and SAAS based LLM's performance on Tabular Medical Trial Data.
• 2. Experimented with different Context Gathering approaches such as RAG, RAG+Reranking, Tabular To Text with prompting, Multi Instruct Prompt, Knowledge Graphs and Langchain Agents specifically SQL Agent.
• 3. Built a Scalable and Modular RAG + Agent based pipeline for extracting important context and building a Modular LLM handler which can handle both Open and Closed Source LLMs.

Jan 2022 – Jul 2023 · 1 yr 6 mos · Bangalore Urban, Karnataka, India

• 1. MLOps Framework: Led a team in creating AWS Pipelines Framework for deploying ML models across the organization.
• 2. MLOps Pipeline: Led team in creating scalable pipeline for SKU classification, reducing training time by over 50% using Pyspark and Google Composer.
• 3. IoT - CV - ML Exploration: Spearheaded development in IoT and computer vision, integrating edge devices with IoT sensors and video streams.
• 4. Anomaly Detection - ML in Cybersecurity: Developed MLOps pipelines for anomaly detection using hybrid models and automated ML lifecycle with GCP services.
• 5. Led the GCP ML Advanced Specialization Partnership from the Machine Learning Front.

Aug 2021 – Jan 2022 · 5 mos · Remote

• 1. Document Text Extraction and Summarization: Created an API for extracting text from PDFs, grouping pages using OpenCV-based logic, block detection, and Tesseract OCR.
• 2. Implemented Hugging Face's T5 Transformer for text summarization of extracted content.
• Explored state-of-the-art NER models and datasets for Named Entity Recognition tasks.
• 3.Extracted information from .docx (XML) formatted documents using traditional Python libraries, integrating with NER and NLP-based tagging pipelines.
• 4. Explored SPARQL and Semantic Web approaches for graph creation to establish content linking.

Aug 2021 – Dec 2021 · 4 mos · Remote

• Retail Analytics Forecast: Creating a Proof of Concept Forecast model for online retail fashion products to assist the planning team with stock intake decisions, utilizing a combination of classification and regression algorithms for solving the problem.

Feb 2021 – Jul 2021 · 5 mos · Bangalore Urban, Karnataka, India

• 1. Computer Vision Pipeline for Targeting and Navigation: Developed traditional and deep learning-based feature matching algorithms (SIFT, SURF, D2 Net, SuperGlue, LoFTR) and SLAM-based algorithms, improving accuracy and generalization.
• 2. Platform Shift - Raspberry Pi to Jetson NX: Led the transition of deployment architecture from Raspberry Pi to CUDA-based platforms (Jetson Xavier NX, Jetson Xavier AGX) for enhanced computational power and accuracy, utilizing OpenVINO, TensorRT, and DeepStream.
• 3. Machine Learning- Common Duties: Collaborated with the ML Engineering team to convert and optimize models into OpenVINO formats, enhanced object detection models, and contributed to model training and deployment.

May 2019 – Feb 2021 · 1 yr 9 mos · Bengaluru Area, India

• 1. Computer Vision for Safety: Developed CV models for theme park safety, including image classification, object localization, tracking, pose estimation, and cloud integration.
• 2. Computer Vision for Suspect Search: Built CV pipeline for suspect identification and anomaly detection in video feeds, focusing on object localization, tracking, person re-identification, search logic, clustering logic, and cloud integration.
• 3. Document Classification with Transformers: Integrated NLP-based transformer models (Open GPT-2, DistilBERT, Roberta, Longformer) into Kubeflow pipeline, evaluating latency and accuracy.
• 4. Translation Service for Document and Web: Developed solution for document and web page translation using Google Cloud APIs (Vision, Translation, Doc Parser) and Auto ML.
• 5. Hybrid Model Deployment Scenarios - Solutions research: Explored NVIDIA DeepStream SDK, TLT, TensorFlow Lite for CPU deployments, and CPU optimizers like Intel MKL-DNN, OpenVino. Explored federated learning for privacy and distributed model development.

Jun 2018 – Sep 2018 · 3 mos · Kozhikode, Kerala, India

• Developed an End-to-End (Edge-Cloud) Object/Number Search Pipeline which can enable the users search for any entity across different video streams hosted on Raspberry Pi's connected to a Centralised Cloud inference system. Used Google Vision API's, Yolov3 as a inference API.

May 2016 – May 2019 · 3 yrs · Kozhikode, Kerala

• 1.Diverse robotics projects: Experience includes developing autonomous bots, integrating vision and sensory models, creating a semi-humanoid robot, and building quadcopters and hexacopters.
• 2. Rehabilitation system: Designed a knee rehabilitation device with a flexo-extension mechanism.