Data Scientist & AI Engineer - Production ML Systems Specialist

Deepiga Loganathamoorthy

+60-143****** . **********@*****.*** . linkedin.com/in/deepigaloganathamoorthy . Github Professional Summary

Data Scientist & AI Engineer with 12+ production deployments, delivering LLM and ML systems with 60–85% efficiency gains. Strong in end-to-end system design, deployment, and responsible AI. Work Experience

Hayat Technologies Sdn Bhd Malaysia

Data Scientist March 2024 – Present

• Led end-to-end development of production RAG/CAG systems including multi-channel chatbots, pol- icy assistants, and conversational agents; owned system architecture design, knowledge base curation, hybrid retrieval implementation (semantic + TF-IDF), and LLM guardrail engineering.

• Designed and deployed ML prediction pipelines for public health surveillance and business intelligence; managed full workflow from feature engineering and model training to FastAPI + MySQL deploy- ment, versioned pipelines, and controlled backfilling.

• Built decision-support platforms integrating real-time analytics, geospatial visualization, and LLM- assisted summarization; delivered production systems using Python, Dash, and API-first architecture serving multiple government agencies.

• Owned data analytics infrastructure including competency measurement frameworks, survey intelligence pipelines, and qualitative-quantitative integration; automated sentiment extraction and thematic analysis from large-scale free-text responses.

• Established MLOps practices including Docker containerization, CI/CD pipelines, environment parity

(dev/staging/prod), and human-in-the-loop validation workflows; ensured deterministic inference and audit- ready outputs.

• Conducted system benchmarking and optimization against baseline retrievers; implemented OCR integration, metadata-aware filtering, and structured output enforcement to improve contextual relevance across multilingual queries (English, Bahasa Melayu).

• Collaborated with cross-functional stakeholders including policymakers, healthcare professionals, ESG con- sultants, and event organizers to translate domain requirements into production AI solutions achieving 60–85% efficiency gains.

AXA / Europ Assistance Malaysia

Junior Business Intelligence (promoted to IT Developer – Data Science) Apr 2023 – Dec 2023

• Led ML Proof of Concept from research to production deployment; owned full pipeline including data preprocessing, feature engineering (one-hot encoding, SMOTE), transfer learning, hyperparameter tun- ing, and model evaluation achieving 93%+ accuracy.

• Integrated ML models into RPA systems for automated medical claim approvals; managed continuous model refinement and threshold optimization post-deployment to maintain 93%+ accuracy with incoming data.

• Developed Power BI dashboards for Finance, HR, and Telephony teams; automated MSSQL and MySQL tasks using advanced SQL (CTEs, complex CASE statements), streamlining reporting processes and de- livering monthly performance presentations.

Education

M.Sc. in Data Science 2023

University of Malaya, Malaysia

• OnSIDe: Suicidal Text Analysis (BERT, CNN, Hybrid BERT–CNN) – Project Lead Project link: https://peerj.com/articles/cs-3239/

• Audio Based Depression Classification (CNN, SVM, Random Forest, and ANN models with acoustic features)

B.Sc. (Hons.) in Biotechnology 2018

AIMST University, Malaysia

Skills

• AI/ML Engineering: Retrieval Augmented Generation (RAG), Hybrid Retrieval (Semantic + TFIDF

+ Rule Based), Agentic LLM Systems, LLM Guardrails & Constrained Generation, Conversational AI, Time Series Forecasting, Supervised Tabular ML, Feature Engineering (Temporal Lags, SMOTE, One-Hot Encoding), Transfer Learning, Probabilistic Forecasting, Model Benchmarking

• LLM & NLP Systems: Prompt Engineering & Orchestration, Guardrails & Constrained Generation, Multi-Turn Dialogue Management, Structured Output Enforcement, Hallucination Mitigation, Multilingual NLP (English,Bahasa Melayu), OCR-Assisted Document Processing, OCR Assisted Ingestion, Metadata Aware Retrieval, Sentiment Analysis, Thematic Extraction

• Backend & Production Systems: FastAPI, API First Design, Stateful Workflows, Deterministic Pipelines, SQL (MySQL, MSSQL), Schema Locked APIs, Versioned Pipelines, Controlled Backfilling, Human in the Loop Systems, Production ML Deployment, RPA Integration

• MLOps & Infrastructure: Docker, CI/CD (GitHub Actions), Environment Parity (Dev/Staging/Prod), Model Versioning, LangChain, ChromaDB, Vector Databases, TensorFlow / PyTorch, Dash, REST APIs

• Data Engineering & Analytics: Python (Pandas, NumPy), Power BI, DAX, Dash, Geospatial Analytics, Cohort Analysis, Qualitative-Quantitative Integration, Sentiment Analysis, Thematic Extraction Selected Projects & Research

• Grant Eligibility Assistant Chatbot(Cik Awan) : Production RAG chatbot across social and web channels reducing grant discovery to minutes and surfacing 3–5 eligible grants per session.

• Hybrid RAG Policy Knowledge Assistant : Multilingual hybrid RAG system using semantic chunking, embeddings, TF-IDF, and OCR with higher contextual relevance than baseline retrieval.

• Context-Adaptive ESG Assessment Agent : CAG-based ESG workflow replacing static surveys with semantic analysis and deterministic schema mapping improving SME data quality.

• Airbo Dengue - ML-Powered Outbreak Prediction Platform : Supervised ML pipeline using lagged climate features delivering probabilistic 2–12 week sublocality-level forecasts for public health stakeholders.

• SEJAGAT - Policy Observatory & Action Monitoring Platform : Production policy observatory integrating action tracking, geospatial analysis, and LLM summarization for real-time multi-agency moni- toring.

• AI-Powered Conference Business Matching System : LLM-based compatibility scoring with human- in-the-loop validation achieving 80–85% stakeholder satisfaction across 15–18 sessions.

• Selangor Care Center Policy Insights Dashboard-LLM Automated & Sentiment Analysis : Automated NLP pipeline analyzing 200+ survey responses extracting sentiment and themes for Power BI dashboards.

• Drug Text Classification for Medical Claims : Production System at AXA Transfer-learning-based drug classification achieving >93% precision and improving RPA efficiency by 60–70%. Awards

• 2025: Semi-finalist, DerivAI Hackathon (JomHack) 2025 – VISIRO Fraud Detection App

• Feb 2023: Finalist, Pharma-Hackathon 2022 – Entrepreneurial Pitch Presentation Leadership & Community Service

Public Relations Officer, Rotary Club of Kuala Lumpur West (2020–2021) Led communications and supported community service initiatives contributing to the Service Above Self Award

(2021).

Founder, Provision (Non-Governmental Organization) Founded a leadership-focused NGO and led student programs, outreach initiatives, and sponsorship acquisition.

*References available upon request.

Contact this candidate