top of page

Section 1: Professional Experience & Work Projects

​

Datastrato.ai | Software & Data Engineering Intern

Jun 2024 – Aug 2024 | San Jose, California

Project: Apache Gravitino Metadata Discovery Integration

​​​​

  • Integrated Apache Gravitino with 5+ diverse ML datasets,

  • Streamlining metadata discovery and reducing onboarding time by 30%.​

  • Optimized CI/CD pipelines using Dockerized test environments

  • Resulting in a 15% boost in deployment reliability.​

  • Authored technical documentation for high-scale data infrastructure.

Vodafone Qatar | Software Engineer (Backend Systems)

Aug 2021 – Mar 2022 | Pune, India

Project: Enterprise Product CRM Workflow Automation

​​​​

  • Architected end-to-end CRM workflows using Bonitasoft BPMN,

  • Automated manual verification processes for thousands of subscribers and Sim Card Users.​

  • Developed robust Java-based backend services and RESTful APIs.

  • Synchronized real-time customer data across legacy systems.​

  • Collaborated with cross-functional teams to deploy updates in an Agile/Scrum environment.

  • Maintained 99.9% system uptime.

Section 2: Research & Academic Projects

​

PixelPerfectAI | Image Restoration SaaS Web App

Tech Stack: ESRGAN, GFPGAN, Colorization, Diffusion, Deep Learning

​​​​​

  • Built a distributed image restoration pipeline using ESRGAN, GFPGAN, and deep learning-based colorization models to enhance low-resolution and degraded images.

  • Developed containerized worker services with asynchronous task queues to enable scalable, reliable, and fault-tolerant image processing.

  • Implemented monitoring and performance tracking to maintain system health, throughput, and operational stability in production environments.

  • Achieved 99.5% job success rate with low median latency, supporting high-volume real-time and batch image enhancement workloads.

  • Optimized inference workflows to process thousands of images efficiently while ensuring consistent output quality and scalability.

Community Connect | Get Help and Receive Help (Mobile)

Tech Stack: FlutterFlow, Firebase, Low Code, Mobile App Development

  • Developed Community Connect, a mobile resource-exchange platform using FlutterFlow and Firebase to facilitate sharing of skills, goods, and services within local communities.

  • Built real-time discovery, multi-category filtering, and structured matching workflows to streamline exchanges from request creation to completion.

  • Engineered a scalable Cloud Firestore backend with role-based access control and atomic state management for secure and reliable transactions.

  • Designed and implemented end-to-end user experiences that foster trust, engagement, and collaboration among students, seniors, and volunteers.

  • Optimized data synchronization and workflow automation to support concurrent exchanges with high reliability and zero data loss.

​

EduLoyalty | Gamified Reward Platform

Tech Stack: TypeScript, React, SQL, Flask

  • Built a full-stack rewards and recognition platform using React, Flask, and SQL, enabling points, badges, coupons, and achievement-based incentives to drive student engagement.

  • Designed and developed an interactive user dashboard showcasing accomplishments such as first-time sign-ups, coupon redemptions, leaderboard rankings, profile highlights, and hall-of-fame placements.

  • Developed scalable RESTful APIs with Flask to manage user activities, rewards distribution, achievement tracking, and transactional workflows efficiently.

  • Engineered a relational SQL database with optimized schema design and queries, ensuring secure, consistent, and high-performance data storage and retrieval.

Emotion-Cause Extraction | NLP Research (SemEval 2024)

Tech Stack: RoBERTa, Hugging Face, Python

  • Detected emotion-cause in sentences using transformer to identify emotional triggers in FRIENDS dataset. 

  • Optimized tokenization and preprocessing workflows, reducing total model training time by 28%.

  • Achieved 74% accuracy, placing the model in a competitive rank against global SemEval benchmarks.

Section 3: Publications (The "Citations" Impact)

​

Multimodal AI & Sensor Fusion for Industrial Safety 

Symbiosis Centre of Applied Artificial Intelligence | Pune, India

Tech Stack : CNN, LSTM, Numpy, Pandas, Sensor Fusion

  • Architected a CNN-LSTM Hybrid Model for Gas Leakage Detection that fused thermal imagery and

  • chemical sensor data, achieving 96% accuracy in real-time gas leak detection and identification.

  • Curated & Published "MultimodalGasData" - a custom dataset (published in MDPI Data) involving

  • manual thermal data collection and 3x data augmentation.

  • Outperformed single-modality industry baselines by 12-18%, providing a fail-safe mechanism.

  •  Co-authored 5 peer-reviewed papers with 210+ total citations, in mulitmodal sensor fusion.

Applied System Innovation (127 Citations)

​​

Soft Computing – Industry 5.0 Tasks (42 Citations)

​

MDPI Data – MultimodalGasData (37 Citations)

​​

Section 4: LLM + Agentic AI Applications

​

MeetingMind — 3 Step Agentic AI LLM Tool

San Carlos, California 

Tech Stack : Claude API , Frontend, Backend

  • Developed MeetingMind, an agentic AI application that converts meeting transcripts into action items, decision summaries, and follow-up emails in under 10 seconds.

  • Engineered a 3-step Claude API workflow to automate transcript analysis, summarization, and email generation with minimal user input.

  • Built a lightweight, zero-backend web application using native Web Speech API for real-time voice transcription and note capture.

  • Designed an end-to-end AI-powered meeting productivity solution that streamlines post-meeting documentation and task tracking.

  • Optimized processing pipelines for fast, reliable generation of structured meeting insights and actionable outcomes.

​

AutoHeal | Clinical AI Dashboard

Beta Fund x Butterbase Hackathon

Tech Stack : Multi Agents, Observability, Sensor Fusion

  • Developed AutoHeal, an AI-driven clinical operations dashboard that monitors patient signals, prioritizes risk events, and automates escalation workflows in real time.

  • Designed a multi-agent decision system leveraging signal fusion, triage scoring, and autonomous action routing to identify and respond to critical patient conditions.

  • Built an interactive operations interface for tracking patient monitoring, alert generation, escalation accuracy, response times, and system performance metrics.

  • Implemented intelligent prioritization logic using time-to-harm, reversibility, and confidence scoring to surface high-risk cases and reduce false alerts.

  • Integrated continuous learning and self-evaluation mechanisms ("Dream Mode") to capture operational insights and improve agent decision-making over time.

bottom of page