Arjun Sasikumar
Profile Image

Arjun Sasikumar

Applied AI researcher and engineer
B.Tech, Electrical and Electronics Engineering, NIT Nagaland · BS, Data Science and Applications, IIT Madras · Bengaluru, India

I work at the intersection of applied AI, software engineering, and industrial systems. I develop machine learning methods for signals and industrial processes, publish research, and build production software for real-world use. My experience spans edge anomaly detection, uncertainty-aware fault diagnosis, process optimization at Yokogawa, and full-stack systems for clients in the UAE. My research focuses on adaptive intelligence in cyber-physical systems, building systems that learn, adapt to changing conditions, and operate reliably in the real world.

9.24B.Tech CGPA, First Class with Distinction
6papers, IEEE and Springer
Best PaperIEEE CCPIS 2025
YokogawaIndustrial AI, Abu Dhabi
3client platforms in production
Emphasis

Education

Two concurrent degrees: Electrical engineering and Data Science.

2022 – 2026

B.Tech, Electrical and Electronics Engineering

National Institute of Technology (NIT) Nagaland · CGPA 9.24/10 · First Class with Distinction

Linear Algebra, Multivariate Calculus, Probability and Random Processes, Data Structures and Algorithms, Signals and Systems, Linear Control Systems, Power Electronics, Embedded Systems, Robotics.

2022 – 2028

BS, Data Science and Applications (online)

Indian Institute of Technology (IIT) Madras · Ongoing

Machine Learning, Deep Learning, Data Structures and Algorithms, Data Analysis, Programming, Tools in Data Science.

Higher secondary

98.8%

Calicut, Kerala

Industry Experience

Most recent first.

2026 – Present

Independent Software Engineering Consultant

Sole engineer, from requirements to production deployment

HireLocal, Nagaland Skill and Entrepreneurship Development Mission (NSEDM) In development

Contracted to build a digital employment platform that connects certified candidates with approved local entities and verifies actual engagement outcomes for programme monitoring. Scope: candidate data integration with the existing MIS (CSV import or API), employment and gig workflows, outcome verification with anomaly flags, role-based access control, and production readiness covering hosting, security audit and data governance.

Noviqa Management Services, Dubai · Jul 2026

Built and maintain the production web platform for a commercial receivables company (Next.js, TypeScript, Tailwind, Vercel). Server-side validation, sanitization, honeypot and rate limiting, CSP and HSTS headers, domain, SSL, DNS and authenticated email.

BlueOcean · 2026

Business management platform integrating accounting and internal operations, plus the company website. Requirements, system design and delivery with stakeholders.

RV Group Contracting, UAE · 2026

Production site for a Civil, MEP and HVAC contractor across four emirates: services, per-emirate licensing, end-to-end quote-request flow, SEO metadata and image delivery.

Dec 2025 – Feb 2026

Advanced Solutions Engineering Intern

Yokogawa, Advanced Solutions and Optimization Department · Abu Dhabi, UAE

Worked with industrial time-series data, process variables, statistical modelling, machine learning, and engineering constraints in real-world process systems.

Ethane cracking optimizationBorouge AI Optimizer
Conversion-prediction models and severity-aware multivariate optimization for safe steam-to-hydrocarbon (S/H) ratio selection.
Refinery soft sensorsENOC
Inferential flash-point models for naphtha, kerosene and gas oil from live process data, using Linear Regression, PCR and PLS with residual diagnostics.
Retrieval over furnace dataProcess RAG
Vector retrieval of similar high-conversion operating clusters, recommending setpoints grounded in the plant’s historically safe envelope.
Well-test automationOil and gas
Python engine for accept/reject analysis with rule-based logic, data-quality checks, time-series visualization and automated reporting, reducing manual review.
Python · SciPy · Scikit-learn · PCR/PLS · FAISS · Pandas · Plotly

Research Experience

Undergraduate Researcher, NIT Nagaland, in collaboration with Dr. D. Ganga · 2025 – Present

Paper [1]

Edge-efficient autoencoders for unsupervised ECG arrhythmia detection Best Paper

Flat VAE and LSTM-VAE trained only on normal cardiac signals, with reconstruction-error thresholding that needs no labelled pathology. A 107 KB model and a High-Channel Shallow Conv1D Autoencoder with global feature summarization.

99% accuracy · 0.97 macro-F1 · 0.995 AUC · 5.35 ms on CPU
Papers [4] [5] [6]

Machine learning for bearing fault diagnosis from vibration signals

Classical ML, BN-augmented Bi-GRU and VAE anomaly detection under varying load and speed. Current work on class-conditional latent-space scoring for selective classification, and an edge-compatible anomaly scoring layer that flags unreliable 1D-CNN predictions under distribution shift.

3 papers published · 2 manuscripts in review
Paper [3]

Real-time computer vision for industrial safety compliance

Customized YOLOv5s for PPE compliance on construction sites, with CLAHE and an occlusion-aware variant. Quantized and exported to ONNX; also benchmarked on TensorFlow Lite and TensorFlow.js.

0.980 mAP@0.5 · 0.95 F1 · ~38.7 FPS, CPU only

Research interests

Focus

Adaptive intelligence in cyber-physical systems

How intelligent systems recognize that their environment has changed, adapt reliably within uncertainty, compute and safety constraints, and stay trustworthy when deployed in the physical world. This builds on my work in anomaly detection, representation learning and uncertainty-aware diagnosis.

Adaptive learning · Uncertainty · Safe adaptation · Edge AI · Cyber-physical systems

Publications

Six conference papers, three as first author. * corresponding author.

  • Lightweight and Efficient Variational Auto Encoder for Arrhythmia Detection Best Paper AwardA. Sasikumar, D. Ganga* · Int. Conf. on Circuits, Power and Intelligent Systems (CCPIS), Bhubaneswar, Sept 2025 · IEEE · DOI
  • Multi Modal Edge Data Processing for Real Time Landslide Early Warning SystemD. Ganga, N. Ngullie, R. Kumar, A. Sasikumar · Int. Conf. on Recent Advances in Computing and Systems (REACS), Gwalior, Dec 2025 · IEEE · DOI
  • Real-Time Detection of Personal Protective Equipment in Construction Sites Using YOLOv5: A Computer Vision-Based Safety Compliance FrameworkA. Sasikumar, D. Ganga*, N. Ngullie · Int. Conf. on Pattern Recognition and Machine Intelligence (PReMI), IIT Delhi, Dec 2025 · Springer LNCS, 2026 · DOI
  • Intelligent Fault Classification of Rolling Bearings via Stacked BN-Augmented Bi-GRUD. Ganga*, A. Sasikumar, S. Kumar · Int. Conf. on Advances in Information Engineering and Systems (ICAIES), Dec 2025 · Springer LNCS
  • Lightweight Classical Machine Learning for Edge-Ready Bearing Fault Diagnosis: A Comparative Analysis of Random Forest and SVM on Statistical FeaturesA. Sasikumar, D. Ganga* · Int. Conf. on Vibration Engineering and Technology of Machinery (VETOMAC), IIT Guwahati, Dec 2025 · Springer LNME
  • Variational Autoencoder-Based Anomaly Detection in Vibration Signals for Early Bearing Fault DiagnosticsD. Ganga*, A. Sasikumar, S. K. Mandal · VETOMAC, IIT Guwahati, Dec 2025 · Springer LNME

Manuscripts in review

  • A One-Shot, Lightweight, High-Channel Shallow Edge-Ready Conv1D Autoencoder with Global Feature Summarization for Early and Efficient Diagnosis of Cardiovascular Disorders
  • Edge Compatible Anomaly Scoring of 1D CNNs with Confidence Quantification for Explainable Fault Classification under Uncertain Conditions
  • Trust-Gated Selective Classification for Bearing Fault Diagnosis via Class-Conditional Latent-Space Scoring
  • Edge-Optimized Multi-Class Threat Detection Using YOLO: An Occlusion-Aware, CLAHE-Enhanced Framework for Real-Time Intelligent Surveillance

Selected Projects

Systems built end to end, from data model to deployment.

Full-stack AI · Finance operations

Trace: AI Finance Controller

Reconciliation platform with an AI investigator, an Apache AGE evidence graph, and a policy engine as the only component allowed to auto-close a case. An independent Verifier checks the evidence chain, and humans approve or reject escalations.

94.4% resolution accuracy · false auto-closes cut from 33.3% to 0%
FastAPI · Postgres · Apache AGE · Docker
Multi-tenant SaaS · Collections

Recoupa: debt recovery and sales CRM

Workspace for debt-collection agencies: calling queue, cheque tracking, ageing-based commissions, disputes, spreadsheet import and an audit trail. Every mutation is re-checked server-side against a permission matrix.

Next.js · Drizzle · Neon Postgres · RBAC
Applied AI · Retrieval-augmented generation

Synapse: RAG health assistant

FAISS retrieval grounding Llama-2-7B in WHO, OpenMed and medical-encyclopedia sources, with symptom-based doctor recommendations and reminders.

Accuracy 65% to 91% · retrieval latency ~500 ms to under 200 ms
LangChain · FastAPI · FAISS
Mobile · Offline-first

Vendora: billing and inventory for local shops

Android app with barcode scanning, employee accounts, cloud backup and sync, built end to end and designed to keep working without a connection.

Kotlin · Jetpack Compose · Room · Supabase
SaaS · Generative AI

Vimra: AI course generation

Course generation platform for educators and content creators, with OpenAI-assisted content, Stripe payments and course management.

Next.js · MongoDB · Prisma · OpenAI · Stripe
Web app · Generative AI

Quizo: AI quiz platform

Responsive quiz application with dynamically generated AI quizzes, scoring and persistent user progress.

Next.js · MongoDB · OpenAI

More repositories at github.com/83Gh0st.

Leadership and Recognition

Awards and service alongside the research and engineering work.

Honors and awards

Aug 2026

First Class with Distinction

B.Tech, Electrical and Electronics Engineering, NIT Nagaland. CGPA 9.24/10.

Nov 2025

First Prize, Hackathon

National Institute of Technology Nagaland.

Sept 2025

Best Paper Award

IEEE 2nd International Conference on Circuits, Power and Intelligent Systems (CCPIS).

2025

Certificate of Recognition

Ministry of Education, Government of India, for contributions to the National Innovation and Entrepreneurship Education Bootcamp.

Leadership

NIT Nagaland

Secretary, Entrepreneurship and Incubation Cell (IIC)

  • Led planning, budgeting and logistics for national-level innovation and entrepreneurship events.
  • Directed multi-tiered student committees and coordinated stakeholders to deliver programs of national importance.
  • Managed operations for the Ministry of Education’s Innovation, Design and Entrepreneurship Bootcamp, training government school principals and teachers across India.
NIT Nagaland

Committee Member, IEEE Student Branch

  • Authored event reports and technical content, and managed communications across web and social platforms.

Skills

 

Languages
Python, TypeScript/JavaScript, SQL, C, Java, Kotlin, MATLAB, Bash
Machine learning
PyTorch, TensorFlow, Scikit-learn, OpenCV, autoencoders and VAEs, CNNs, RNNs, Transformers, anomaly detection, uncertainty estimation, representation learning, edge AI
LLM systems
RAG, AI agents, LangChain, LangGraph, vector databases, Hugging Face, model evaluation and serving
Industrial and process
Soft sensors, PCR/PLS, process optimization, time series, signal processing, control systems, PID, EnergyPlus
Software
Next.js, React, Node.js, FastAPI, Postgres, MongoDB, Prisma, Drizzle, Tailwind, Stripe, Jetpack Compose
Cloud and embedded
AWS, Docker, Vercel, Git, OpenTelemetry, Raspberry Pi, ESP32, Jetson Nano

Contact

Open to research positions and PhD programs, applied AI/ML and software engineering roles, forward deployed and solutions engineering, and consulting or product work.

Emailarjun3314508@gmail.com
Phone (IN)+91 85479 85276
Phone (UAE)+971 56 211 6482
GitHubgithub.com/83Gh0st
LinkedInarjun-s-a884b8258
ORCID0009-0008-7897-6800