
Abu Hasnat Abdullah_
Senior AI Engineer
LLM · RAG · MLOps · Backend
Experience
Senior Software Engineer
@Cefalo Bangladesh Ltd. · Sensa AS, Norway
Own AI engineering for Sensa, a Norwegian industrial-AI company — delivering LLM applications, computer-vision systems, time-series MLOps, and the underlying data platform end-to-end, with deep focus on evaluation, observability, and responsible-AI guardrails.
Senior Software Engineer
@Strativ AB
Led a cross-functional team of 11 engineers across 6 concurrent enterprise projects for top-tier Swedish clients — owning system design, backend architecture (Python, Django), code review, and delivery, while staying hands-on with API and AI feature development.
Software Engineer
@Circle Fintech Ltd.
Shipped Boarding Bay — an eKYC platform for Jamuna Bank — modernizing customer onboarding and accelerating business financial transaction processing across branches. Built 3 production-grade microservices (NID OCR, face detection, billing) that automated identity verification and significantly reduced manual onboarding time.
Junior Research Analyst
@Economic Research Group
Analyzed mobile-app data for Romoni (on-demand beauty marketplace) to drive sales growth and forecast demand using regression models, informing pricing and marketing decisions.
Research Assistant
@North South University
Built the-Waves web application (the-waves.org) to publish research articles; contributed research on technology, innovation, and policy topics.
Founder & CEO
@Shikhte Chai
Founded and scaled an educational initiative in Bangladesh, leading strategy, operations, and program design to upskill students in technology and entrepreneurship.
Projects
Featured Projects
Unity AI
LLM-powered conversational assistant that lets industrial operators query plant data in natural language. Designed a custom RAG pipeline combining a Neo4j knowledge graph with an InfluxDB time-series store, with grounded citations and data-governance rules that drive confident, source-backed answers and minimize hallucinations.
CV Analyser
AI résumé-parsing platform that extracts text from CVs and returns structured, ATS-ready output with role/skill categorization, plus a chatbot for recruiter Q&A over the parsed corpus. Fine-tuned Llama 3.2, DeepSeek, and Qwen with Unsloth (QLoRA), benchmarked trade-offs across model families, and reached 97% accuracy on a local 3B model — matching cloud-API quality at a fraction of the cost.
Fish Food Pellet Detection
Productionized a YOLOv11-nano computer-vision model that detects and counts fish-food pellets in real-time aquaculture footage, lifting detection accuracy from 40% to 98% and unlocking automated feed-rate optimization for sensor-equipped fish farms.
Sensa Engine
Built Sensa's internal data-orchestration and feature platform — the foundation powering Unity AI, Unity-MLOps, and Pellet Detection. Designed Mage.ai ETL pipelines to ingest sensor and external data into time-series and feature stores, with reusable transforms that serve both real-time anomaly detection and downstream RAG retrieval.
Unity-MLOps
End-to-end anomaly-detection pipeline on time-series sensor data with automated training, model registry, deployment, and drift monitoring — turning manual log review into proactive, real-time alerts.
TourGPT
A SaaS RAG chatbot grounded in a custom tour knowledge base, delivering personalized recommendations and end-to-end order management. Owned prompt engineering, embedding strategy, and safety guardrails from prototype to production.
Boarding Bay
3 production-grade microservices (NID OCR, face detection, billing) for Jamuna Bank's eKYC platform — automating identity verification and significantly reducing manual onboarding time.
Other Projects
- Investment Management SystemDjangoREST APIPostgreSQLSaaSKafkaAWS
- Health CheckDjangoREST APIPostgreSQLKafkaNumPyAWS
- Smart College Management with Biometric AttendanceDjangoREST APIPostgreSQLCeleryRedisAWSIOT
- Open Organization — Transparent & Traceable Donation RaiseDjangoPostgreSQL
- KO — Bangla Programming LanguagePythonCompiler Design
- Terrorist Attack Success PredictionPythonPandasscikit-learnMatplotlibSVMRandom ForestANNXGBoost
Certifications
Data Science
Stanford University (Prof. Jennifer Widom)
2024
Business Analysis & Prototyping
LICT (Bangladesh Govt.) & BracIT
2019
Research & Publications
Predicting the Result of a Cricket Match by Applying Data Mining Techniques
2020
Applied Recursive Feature Elimination, ZeroR, Decision Tree, Random Forest, and XGBoost for cricket match prediction.
Latest Blog Posts
Loading posts...