Hello! I Am MD Asif Khan Tonay
Hello! I Am MD Asif Khan Tonay
A Systems Architect who
Judges AI
by its
architecture
...
Because if the architecture does not scale to production, what else can?
I'm a AI Systems Architect|
Currently, I'm an AI Engineer & ML Systems Architect at Betopia Global,
Specializing in Production LLMs, vLLM Clusters, Multi-Agent RAG Orchestration, and Quantitative MLOps. Bridging state-of-the-art research into resilient, low-latency enterprise reality with zero hallucination.
PRODUCTION EXPERIENCE • INDUSTRY
Work Experience
Proven track record engineering scalable AI runtimes, autonomous agent pipelines, and enterprise microservices.
Betopia Limited
R&D Division • AI Systems
Software Engineer (Level 1 - AI Systems)
- check_circle Architecting high-scale LLM workflows, self-hosted vLLM engine inference, and Qdrant vector clustering.
- check_circle Engineered live sales agent automation overseeing 500+ commercial accounts with automated tool calling and Claude integration.
- check_circle Built real-time telemetry pipelines reducing latency by 45% using async NATS streaming and Redis brokers.
Sparktech Agency
Machine Intelligence Unit
Junior AI Developer
- check_circle Developed computer vision pipelines for document OCR, layout decomposition, and preprocessing.
- check_circle Created automated RESTful microservices with FastAPI for batch image classification and entity extraction.
- check_circle Implemented Dockerized continuous integration testing suites across cloud deployment targets.
I'm currently architecting scalable AI systems for production LLMs & multi-agent clusters
delivering sub-second latency, deterministic tool calling, and high-concurrency enterprise pipelines
End-to-end production AI applications architected and deployed in enterprise setups.
Cross-profile AI agents running real-time quotation and negotiation pipelines.
Published deep learning and medical imaging research in IEEE & Springer Nature.
Awarded Galactic Problem Solver in global NASA Space Apps Challenge 2024.
Core Technical Stack & Tooling
Inference Runtimes & LLMs
P99 latency < 20ms • Multi-GPU Tensor Parallel
Vision, OCR & Deep Learning
Segmentation • Feature Pyramids • Spatial AI
Distributed Backend & Messaging
Zero-copy pub/sub • Async event brokers
Data Architecture & ERP Ops
Vector extensions • 2-opt vehicle routing
PRODUCTION SYSTEMS • LAB
Architected & Shipped Systems
100% Authentic Enterprise Implementations • Monitored Under Active Production Scale
Featured Project • Autonomous Workflows
Sales Desk — Sales Intelligence Agent
Enterprise autonomous agent pipeline capable of live buyer intention parsing, dynamic quotation assembly, and price negotiation across 500+ distributed commercial accounts. Implements an automatic multi-tier fallback chain (self-hosted vLLM to cloud endpoints) to guarantee 99.9% uptime.
Featured Project • Enterprise RAG
Betopia Limited AI Chatbot & Retrieval Core
Enterprise customer intelligence agent integrating hybrid dense/sparse vector search with Qdrant. Built with 2-tier role-based access control (RBAC), automated human agent handoff triggers, and OpenAPI compliance across sensitive commercial knowledge bases.
Featured Project • Algorithmic MLOps
AkBORIA SR — AI Field Service Routing System
Real-time Odoo ERP automated sync engine handling field rep routing with a nightly 6-parameter shop priority ML scoring model. Deployed custom OSRM routing heuristics (Nearest-Neighbour, 2-opt, Or-opt) and native Bengali NLP parsing for field order intake.
Featured Project • Mobile Agent Ecosystem
Task Mama — Everyday AI Companion & Micro-Delegation
Automated task intelligence and micro-delegation mobile agent ecosystem. Features voice-first cognitive unload ("Stop carrying everything in your head"), adaptive daily agenda triage, automated calendar and meal planning, and multi-profile family task synchronization.
Featured Project • Adaptive EdTech AI
LearninGPT — AI-Powered Smart Learning Engine
Curriculum-adaptive educational synthesis agent using fine-tuned open models for personalized STEM conceptual tutoring and automated evaluation. Powers 2,871+ interactive AI activities with individualized student coaching.
Education & Degrees
M.Sc. in Computer Science & Engineering
Jahangirnagar University (JU)
Faculty of Mathematical & Physical Sciences • Savar, Dhaka
Advanced graduate curriculum centered on distributed systems architecture, advanced machine learning paradigms, vector store index heuristics, and high-throughput data pipelines.
B.Sc. in Computer Science & Engineering
East West University (EWU)
Department of CSE • Aftabnagar, Dhaka
Major in Intelligent Computing & Algorithmic Foundations. Graduated with honors, distinguished capstone design award in computer vision-based healthcare telemetry, and comprehensive foundations in data structures and algorithmic complexity.
Research Publications
Early Detection of Laryngeal Cancer from Endoscopic Imagery Utilizing Attention-Guided Deep Convolutional Networks
Conducted specialized transfer learning benchmark achieving 98.7% validation accuracy on clinical high-resolution endoscopic datasets with Grad-CAM interpretability and dual attention maps.
Automated Gastrointestinal Tract Abnormality Segmentation Using Multiscale Feature Pyramid Architectures
Trained dual-encoder segmentor for real-time video endoscopy frame streaming with low parameter overhead and sub-15ms edge inference.
Stacked Ensemble Meta-Learning for Remaining Useful Life (RUL) Prediction in Lithium-Ion Energy Cells
Architected gradient boosted ensemble with temporal cross-validation, improving RUL error bounds by 24% under variable ambient thermal stress.
Honors, Fellowships & Recognitions
Global recognition for computational satellite telemetry pipelines and planetary anomaly localization algorithms developed under NASA Space Apps.
Selected for advanced prompt-caching systems, Claude 5.5 Opus architectures, agentic tool-use SDKs, and deep reasoning workflows.
Secured top placements across nationwide machine learning and algorithmic engineering competitions for zero-shot computer vision and real-time distributed routing systems.
