Skip to content
AI Engineer / ResearcherChennai, India — 2026

NavaneethSivakumar.

I ship production AI for the maritime industry at Exaqube — working with the CTOs and CXOs of global shipping companies. First author of CRAFT in Nature Scientific Reports (2026) and SURE in IEEE Access (2025).

Resume Email
Role
AI Engineer, Exaqube
Research
Nature SR ’26 · IEEE Access ’25
Award
Bengaluru Last Mile ’25
Focus
LLMs · RecSys · Vision
05+
Projects
02
Papers
01+
Years
Navaneeth Sivakumar — AI Engineer at Exaqube Technologies
AI ENGINEER/LARGE LANGUAGE MODELS/RECOMMENDER SYSTEMS/FEDERATED LEARNING/COMPUTER VISION/NATURE · SCIENTIFIC REPORTS/IEEE ACCESS/MARITIME AI/
01Profile

About

I'm an AI Engineer at Exaqube Technologies and a fresh graduate of VIT Chennai (B.Tech CSE — AI/ML, CGPA 9.03/10). My favorite kind of work is the part where a half-baked research idea grows up into something an executive actually opens on Monday morning.

At Exaqube I build AI for the maritime industry — tools that turn the chaos of global shipping operations into clear, fast answers for the people running the business. On the research side, I'm first author of CRAFT in Nature Scientific Reports (2026) and SURE in IEEE Access (2025), I won the Bengaluru Last Mile Challenge 2025, and I helped build a patent-pending clinical LLM workflow at Siemens Healthineers.

Outside work I read too many papers, break models on purpose, and hunt for the next problem worth chasing.

Languages

PythonC++JavaSQLTypeScriptHTML/CSS

ML & AI

PyTorchTensorFlowScikit-LearnComputer VisionNLPLLMs

Tools

FastAPIOpenAI SDKNext.jsTailwindGitDockerLinux
Photograph of Navaneeth Sivakumar, AI Engineer at Exaqube Technologies
Fig.02 — Navaneeth Sivakumar
02Work

Experience

Exaqube Technologies logo
Jul 2025 — Present

Built AI-powered systems for the maritime industry, focusing on voyage cost optimization, revenue prediction, and operational intelligence.

Problem

Executives in large shipping companies struggled with delayed access to critical business insights. Data retrieval depended heavily on manual report generation, causing significant delays in decision-making.

Solution

Developed AI-powered platforms enabling real-time querying, analysis, and explanation of enterprise data. Transformed raw operational data into actionable insights within seconds.

Key Impact

  • —Enabled top-level executives to access business-critical insights instantly.
  • —Eliminated dependency on manual reporting workflows.
  • —Built systems allowing drill-down from high-level summaries to granular cargo-level data.

“Regularly interacted with CTOs, CFOs, and CXOs of major shipping companies to understand their decision-making challenges and translate them into AI-driven solutions.”

Siemens Healthineers, Bangalore logo
Jan 2025 — Jun 2025
  • —Designed and developed a PoC for an LLM-powered clinical application to improve healthcare workflow efficiency; the solution is currently in the process of patent filing.
  • —Designed and implemented a FastAPI server architecture to unify access to OpenAI SDK and locally hosted LLMs via OpenAPI endpoints.
Intel logo

Intel Unnati Industrial Training Program

Intel
Jun 2024 — Jul 2024
  • —Developed a system to analyze contract documents using the Contract Understanding Atticus Dataset (CUAD).
  • —Incorporated text preprocessing and Named Entity Recognition (NER) for clause classification.
  • —Integrated template comparison to detect and highlight deviations, ensuring contract compliance and consistency.
Indian Institute of Science (IISc), Bangalore logo
Aug 2023 — Apr 2024
  • —Worked on advanced machine learning projects and research in computer vision.
  • —Collaborated with research teams on cutting-edge AI applications.
  • —Developed and implemented novel algorithms for computer vision tasks.
03Built

Products

Production enterprise AI for maritime operations, shipped at Exaqube.

01

Executive Dashboards

AI-enhanced dashboards built specifically for top-level executives, providing rapid, reliable insights from boardroom summaries to granular cargo-level traceability in seconds.

Executive IntelligenceData Analytics
02

DataSense

AI-powered database querying system that answers questions and explains WHY using historical data and ERP deviation logs.

NLPExplainable AIEnterprise AI
03

QubeSense

Document intelligence platform that parses, organizes, and answers queries from enterprise documents with structured storage.

Document AIKnowledge Systems
04

SailWithAI

AI-driven CRM that recommends customer engagement strategies based on historical interactions, schedules, and location. Features real-time sync with HubSpot.

CRM AIRecommendation Systems
04Award

Achievements

Hackathon Winner

Bengaluru Last Mile Challenge 2025

Winner of the Bengaluru Last Mile Challenge 2025, focused on solving urban mobility using AI.

Complex Problem

  • —Predicting bus & auto travel times across multi-modal journeys.
  • —Handling noisy and missing transport datasets in real-time.
  • —Estimating uncertainty in urban traffic and vehicle availability.

Technical Approach

Time-series modelingProbabilistic travel estimationMulti-modal route optimizationData cleaning & imputation
Winning team — Bengaluru Last Mile Challenge 2025
View Project
Night sky over a lone tree — photographed by Navaneeth Sivakumar
Photographed by Navaneeth — Chennai
05Academics

Education

B.Tech in Computer Science

2021 — 2025

Vellore Institute of Technology, Chennai

Specialization in Artificial Intelligence and Machine Learning

CGPA 9.03 / 10.00

Senior Secondary Education

2019 — 2021

Devi Academy Senior Secondary School

Score 92.2%

Secondary Education

2017 — 2019

Narayana Olympiad CBSE School

Score 91.6%

06Research

Publications

Peer-reviewed in Nature Portfolio & IEEE — federated learning, sequential recommendation, cold-start personalization.

★ Featured · Nature · Scientific Reports2026 · First author

CRAFT: Cold-Start Recommender with Attention and Federated Training

Sivakumar, N., John, R.S., Bijo, A. et al.

Published 14 Apr 2026 · DOI 10.1038/s41598-026-47175-5

Federated-learning recommender that tackles the cold-start problem without leaking user data. CRAFT uses an attention mechanism to highlight salient user-item interaction patterns and aggregates per-client updates via Federated Averaging (FedAvg), with NVFlare for distributed deployment. Across MovieLens 1M, Amazon Movies & TV, and CiteULike, CRAFT improves cold-start nDCG@20 by up to 16.8% over state-of-the-art federated baselines such as FedMF and FedGN while preserving privacy.

Federated LearningCold-StartAttentionFedAvgNVFlare
View Publication DOI
IEEE Access2025 · First author

SURE: Session-Based Uninteresting Item Removal for Enhanced Recommendations

Sivakumar, N., Motha, A., Suganeshwari, G., Syed Ibrahim, S. P., Sugumaran, V.

Vol. 13, pp. 43904–43918 · Published 07 Mar 2025 · DOI 10.1109/ACCESS.2025.3549133

Sequential recommendation systems struggle with short user-interaction sequences, often resulting in poor accuracy due to insufficient or uninteresting data. SURE combines association rule mining with backward prediction to enrich sequences and remove irrelevant items, lifting Mean Reciprocal Rank by 7.31% on average over state-of-the-art baselines across multiple real-world datasets.

Sequential RecommendationAssociation Rule MiningMRR +7.31%
View Publication DOI
07Open Source

Projects

MedLensAI — Patient-Centric Diagnostic Assistant

2025
  • —Developed a multi-modal clinical assistant using Google's MedGemma model.
  • —Implemented image-based diagnostic suggestions and cross-referenced report summaries.
  • —Built features for symptom-image correlation and follow-up investigation suggestions.
  • —Created a Streamlit-based interface with image upload and doctor-style prompt box.
View on GitHub

Smart Glasses for Visual Impairment

2024
  • —Developed a smart glass prototype with facial recognition and threat detection capabilities.
  • —Implemented real-time object detection and navigation assistance.
  • —Integrated voice feedback system for enhanced user experience.
View on GitHub

AI-Powered Spam Detection System

2024
  • —Implemented an AI-powered spam detection system for instant messaging platforms.
  • —Developed custom NLP models for message classification achieving 95% accuracy.
View on GitHub

Intelligent Alarm System

2023
  • —Created a smart alarm system using calendar and sleep data for optimal wake-up times.
  • —Implemented machine learning algorithms for sleep pattern analysis.
View on GitHub
08Off the clock

Photography

When I'm not shipping models, I'm chasing light. Every frame here is mine.

Under the stars — photographed by Navaneeth Sivakumar
Under the stars01
Night & branches — photographed by Navaneeth Sivakumar
Night & branches02
Golden hour, Marina — photographed by Navaneeth Sivakumar
Golden hour, Marina03
09Contact

Get in touch

Open to AI/ML collaborations, research, and opportunities.

Interested in collaborating on AI/ML projects or discussing opportunities? I'm always open to connecting with fellow developers and researchers.

Emailnavaneeth.sivakumar66@gmail.comPhone+91 9176359890
LocationChennai, Tamil Nadu, India