Building transparent, deployable AI — from Q1 medical-imaging research to multilingual RAG assistants for education.
Sunzil Khandaker
AI Researcher & Engineer
7
Research outputs
Published & under review
2+
Published papers
Q1 Elsevier journals
Q1
Elsevier journals
Peer-reviewed, indexed
2+
Years in AI research
CV, XAI, LLM/RAG
AI Researcher and Engineer at Daffodil International University (top-5 in Bangladesh). Author of 7 research outputs spanning deep learning, Explainable AI, LLM engineering and IoT, including peer-reviewed work in Q1 Elsevier journals. My research centers on trustworthy computer vision and AI for education — with a parallel track applying these methods to real-world automation systems, and as Co-founder of AbSun Studio.
Hybrid CNN–ViT architectures for fine-grained recognition, paired with diagnostic evaluation frameworks that expose hidden bias and performance plateaus before deployment.
SHAP, LIME and GradCAM-based interpretability, including a novel study of explainability drift introduced by INT8 quantization on edge-deployed models.
Multilingual retrieval-augmented assistants for university learning-management systems, with bilingual evaluation protocols for factual accuracy, safety and cultural appropriateness.
Low-cost sensing and localization systems — from soil-health IoT to coverage-preserving ensemble methods for robust indoor localization.
Peer-reviewed journals, conference papers and a curated dataset spanning deep learning, explainable AI, LLM/RAG and IoT.
Smart Agriculture Technology
A hybrid CNN–ViT model for medicinal plant recognition coupled with explainability, delivering transparent, high-accuracy classification.
Scientific Reports
A novel framework to quantify and mitigate explainability drift introduced by INT8 quantization in edge-deployed vision models. Preprint available.
Computers and Electronics in Agriculture
An explainable hybrid attention network optimized for efficient, edge-based plant disease diagnosis.
Data in Brief
A curated, augmented image dataset for medicinal-plant computer vision. Published in Elsevier's Data in Brief — a Q1 Artificial Intelligence in Agriculture-adjacent venue.
International Journal of Activity and Behavior Computing
IEEE ICCIT-2025
IEEE QPAIN-2026
Flagship projects where research meets engineering — multilingual RAG for education, AI content generation, and explainability under deployment constraints.
Agentic, AI-powered learning & teaching assistant for university LMS
Published · IEEE Access in progress
A multilingual, retrieval-augmented assistant that ingests complex academic PDFs into optimized embeddings for local LLM retrieval, with bilingual (English/Bengali) evaluation protocols for factual accuracy, safety and cultural appropriateness.
Paper: BanglaRAG: Building a Multilingual Knowledge Backbone for LMS (IEEE QPAIN-2026)
View on GitHubAI-powered education content generation, full-stack web application
Active
A collaborative full-stack platform for AI-assisted educational content generation, built with Next.js (App Router) and Drizzle ORM with a component-driven UI and real-time context management.
Quantization impact analysis on explainability
Under review (Q1) · Preprint available
A framework to quantify and mitigate explainability drift introduced by INT8 quantization in edge-deployed visual classification models.
Top-5 University in Bangladesh
B.Sc. in Computer Science & Engineering
3.97 / 4.00
CGPA
Focus: Data Science · Computer Vision · Deep Learning
NanoBio Research Lab, Daffodil International University
AI & Deep Learning
LLM & RAG Engineering
Explainability (XAI)
Automation & APIs
Data & Languages
Linguistics
Alongside academic work at the NanoBio Research Lab, I build production AI systems — voice agents, lead-generation pipelines and operational automation — for service businesses, both inside industry teams and as Co-founder of AbSun Studio.
In a company
Jr. AI Developer · Remote
Co-founded & led
As Co-founder of AbSun Studio, I design and ship production AI systems — voice agents, lead-generation pipelines and operational dashboards — for service businesses (cleaning, HVAC, property), built on n8n, Vapi, OpenAI and FastAPI. Research-grade rigour applied to real business problems.
Visit absunstudio.comInbound/outbound voice agents that qualify callers, book jobs and log calls in real time — wired into scheduling and CRM with automatic knowledge-gap capture.
Multi-channel intake (WhatsApp, voice, web) with AI cleaning, intent routing, escalation alerts and instant replies that turn enquiries into booked work.
Scheduled reporting flows that read bookings, calculate daily revenue and dispatch owner-ready summaries automatically.
Webhook → Extract → AI Agent → Route → Book → Notify
Vapi Webhook → Classify → Book / Log → Knowledge-gap capture
Cron (8PM, Mon–Sat) → Read → Calculate → Send
Production systems built and deployed at AbSun Studio — real client engagements across cleaning, property tech and service automation.
📍 Dubai, UAEWorkflow
WhatsApp → Sara AI → Conflict Check → Book → 8PM Dashboard
📍 United StatesWorkflow
Vapi Voice Call → Extract Lead → Google Sheets → SMS + Email Alert
📍 United States & CanadaWorkflow
Voice Call → Intent Classification → Booking / Escalation → CRM Log
“Before this system, we were losing bookings every night. Now Sara handles everything — the customers don't even realize they're talking to AI. Our response time went from hours to minutes.
“ParrotScout changed how we handle inbound calls. Every lead is captured, logged, and I get a text the second a call ends. It's like having a 24/7 receptionist that never misses anything.
“The Dream Specialist AI handles our intake across two countries without us lifting a finger. It knows exactly what each caller needs and routes them perfectly. It's transformed how we run the practice.
Beyond research and industry, I teach — translating advanced AI engineering into hands-on learning for large cohorts.
Teaching Assistant — AI Bootcamp
May 2026 — Present
Anonymous feedback from learners across Ostad's AI programmes.
“The n8n and RAG sessions completely changed how I think about AI. I went from zero automation experience to building my own WhatsApp AI agent in one week.
“Sunzil has a rare ability to make complex topics feel approachable. The way he explained prompt engineering and memory management was genuinely the clearest I've heard it.
“I was stuck on my LLM fine-tuning project for days. One session with Sunzil and the blockers were gone. He doesn't just fix bugs — he teaches you how to think through them.
“Honestly the best TA I've had. He shows up prepared, explains with real examples from his own research, and actually cares whether we understand or not.
Contact