About Me
Founder · Director · Founding Engineer · Software & AI Engineer
Contact Information
Professional Overview
AI-focused software engineer based in the Greater Toronto Area, with 8+ years shipping production systems end-to-end and a decade of entrepreneurship — citation-grounded RAG, multi-agent LLM systems, vision models, and the AWS infrastructure beneath them. Founding Engineer at Case Intelligence Assistant — air-gapped, fully cited AI for investigative case-file analysis; Director & Founder of Global Crown Digital Corporation and creator of BlindSense — camera-free elderly wellness monitoring with no cameras, wearables, or microphones; previously Senior Software Engineer | Cloud & AI Specialist at Virtual Health Hub, building production-grade telemedicine platforms.
AWS Certified Machine Learning Engineer – Associate and NVIDIA DLI certified in Generative AI with Diffusion Models and Accelerated Computing with CUDA Python, combining deep full-stack engineering expertise with hands-on machine learning and data engineering capabilities.
My work spans the full technology lifecycle — from designing cloud architectures and embedded firmware to training custom computer vision models, serving self-hosted LLMs, and deploying multi-agent AI systems in production. I independently architect and ship entire platforms: an offline investigative AI system on an NVIDIA DGX Spark, an edge-to-cloud IoT sensing product, and healthcare platforms featuring agentic AI with multi-modal interaction, ambient summarization, clinical risk analytics, and custom-trained vision transformers for classification and semantic segmentation.
Key strengths:
- Built production RAG pipelines where every answer cites its exact source — BGE-M3 embeddings, Qdrant vector retrieval, cross-encoder re-ranking, and self-hosted LLM serving on vLLM with speculative decoding and KV/prefix caching
- Designed and productionized multi-agent AI systems (Amazon Bedrock, LangGraph) with structured memory, context injection, streaming I/O, prompt-injection defense, and LLM-as-judge evaluation
- Built end-to-end computer vision pipelines — DINOv3/ViT transfer learning for classification and pixel-level segmentation — deployed on SageMaker with scale-to-zero serverless inference
- Integrated AI deeply into full-stack applications including NLP, ASR/TTS, medical terminology extraction, OCR, diarization, and real-time video/audio analysis
- Engineered edge-to-cloud IoT systems: ESP-IDF firmware on ESP32, WiFi CSI sensing, 60 GHz mmWave radar, and event-driven AWS backends in CDK with Rust Lambda services
- Security and identity engineering: OAuth 2.0/OIDC, RBAC, TOTP MFA, YubiHSM key custody, and tamper-evident audit logging
Beyond engineering – first author of a REB-approved safety evaluation of agentic AI in mental health (4,500+ scored interactions validated against expert clinical review, manuscript prepared for Mayo Clinic Proceedings: Digital Health), bridging applied technology with academic contribution. Driven by building technology that directly impacts care and safety in communities with limited access to specialists.
Areas of Expertise
Programming Languages
Frontend Development
Backend Development
Databases & Storage
Cloud & DevOps (AWS)
Generative AI & Agents
ML & Computer Vision
NLP, Speech & Medical AI
Data Engineering
Embedded & IoT
Security & Identity
AI Workflow Integration
Designed and productionized citation-grounded RAG, multi-agent, and computer-vision pipelines across public-safety and healthcare products. For CiA, built a fully offline RAG pipeline where every answer cites its exact source (page, video frame, transcript timestamp): BGE-M3 → Qdrant ACORN retrieval → cross-encoder re-rank → Gemma 4 31B on vLLM streamed over SSE, with retrieved evidence treated as untrusted to defend against prompt injection. For LifeLine, built an agentic system on Amazon Bedrock (chatbot, medical-scribe, and risk-analytics agents) with streaming I/O, Polly TTS, and Transcribe STT; integrated Chime SDK for sessions and ambient listening to generate real-time summaries and risk signals, with structured memory and secure backend context injection. For SkinScan, trained a unified DINOv3 ViT-L/16 for 10-class wound classification and pixel-level segmentation in one forward pass, deployed on SageMaker async endpoints with scale-to-zero autoscaling, plus Bedrock agents for staging and clinical recommendations and Comprehend Medical for ICD-10/RxNorm inference. All solutions run in isolated environments — air-gapped on-prem or VPC-isolated AWS — with full audit/reporting for privacy and accountability.
Experience Highlights
Case Intelligence Assistant Inc.
Founding Engineer
- Sole engineer of CiA — an air-gapped AI investigative file-analysis platform for law enforcement, fully offline on an NVIDIA DGX Spark — from first commit to CEO-signed V1.0 in under 3 months
- Built a production RAG pipeline where every answer cites its exact source: BGE-M3 → Qdrant ACORN retrieval → cross-encoder re-rank → Gemma 4 31B on vLLM streamed over SSE, with prompt-injection defense
- Cut large-case, cross-document inconsistency analysis from ~6 hours to ~1 hour through vLLM speculative decoding, KV/prefix caching, and entity-graph pruning
- Multimodal ingestion of 13 document formats plus audio/video: Docling/Tesseract OCR, Parakeet ASR with Whisper fallback, NeMo diarization & voice re-ID, zero-shot NER (GLiNER)
- Court-defensible by design: RBAC + TOTP MFA, YubiHSM 2 key custody, hash-chained tamper-evident audit log, Crown-disclosure export, EN/FR bilingual UI
Global Crown Digital Corporation
Director & Founder
- Founded and operate a Canadian software corporation delivering AI, cloud, and IoT product engineering; conceived and built BlindSense end-to-end as sole engineer
- BlindSense: camera-free elderly wellness monitoring fusing WiFi CSI sensing with 60 GHz mmWave radar — no cameras, wearables, or microphones
- ESP-IDF C firmware across a 3-node ESP32 kit — real-time WiFi-CSI capture and multi-sensor fusion at the edge, OTA A/B updates, per-device X.509
- Event-driven AWS backend in CDK: IoT Core MQTT → Rust Lambda services → DynamoDB/SNS/EventBridge; a deterministic stats engine decides, Amazon Bedrock only narrates pre-computed flags with claim-lint guardrails
- Shipped a Next.js 16 real-time dashboard (SigV4-signed MQTT over WebSocket, 3D radar point cloud) and a SwiftUI iOS app with APNs push
Virtual Health Hub
Senior Software Engineer | Cloud & AI Specialist
- Managed the LifeLine project end-to-end — development, cloud infrastructure, and solution architecture — building a multi-agent AI system on Amazon Bedrock (chatbot, medical-scribe, and risk-analytics agents) with RAG, voice via Polly/Transcribe, and Chime SDK video with ambient-listening summarization; HIPAA-aligned VPC
- Developed native Swift/Kotlin Expo modules bridging Chime SDK — published as react-native-chime-sdk, an open-source TurboModule/Fabric library used in the production app
- Trained a unified DINOv3 ViT-L/16 for 10-class wound classification + pixel-level segmentation in one forward pass for SkinScan, deployed on SageMaker async endpoints with scale-to-zero autoscaling and AprilTag calibration for cm²-accurate measurements
- Shipped VHH Command Center — a control-room portal monitoring telehealth deployments across 11 northern Saskatchewan communities with live geospatial fleet map, telemetry alerts, and WebRTC voice/video ring-out
- Integrated Apple HealthKit / Health Connect (30+ biometrics auto-syncing to the care team) and delivered clinician portals in Vue/Quasar + Next.js
- First author of a REB-approved safety evaluation of the platform’s agentic AI — 4,500+ scored interactions validated against expert clinical review — manuscript prepared for Mayo Clinic Proceedings: Digital Health
Virtual Health Hub
Software Developer
- Developed full-stack healthcare applications using React.js, Vue.js, and Node.js
- Implemented AI integrations and cloud infrastructure
- Built dynamic appointment booking and medication management systems with analytics and notifications
- Integrated Amazon Chime SDK with React Native using Swift & Kotlin
Saaska Software Inc.
Software Engineer
- Developed esiKidz childcare management software using React.js and React Native at the enterprise level
- Learned data engineering by creating apps for web crawling, data extraction, and manipulation
- Experienced AWS cloud infrastructure maintenance, including Lambda, Amazon S3, API Gateway
- Performed optimization of developer toolchain, fully automating deployment and minimizing development friction
SELISE Digital Platforms
Software Engineer
- Developed Sunrise-club event booking site for Telco (Sunrise) users of Switzerland
- Implemented OAuth 2 + OIDC, Storyblok CMS, and MongoDB database
- Maintained URL redirect with WAF in production and web firewall security
- Constructed a complete language translation system
Quantic Dynamics Ltd.
Web Developer
- Built fully responsive web applications using React.js, Next.js, and Node.js with MongoDB and Firebase
- Architected frontend and backend solutions with RESTful API integration and asynchronous client-server communication
- Implemented data manipulation, search, and sort functionalities using Node.js
Certifications
AWS Certified Machine Learning Engineer – Associate
Validates expertise in building, training, tuning, and deploying machine learning models using AWS services.
Verify on Credly →Generative AI with Diffusion Models
June 2026 · Certificate ID: tZ_6rIcjRReMksD-FZbAOw
Fundamentals of Accelerated Computing with CUDA Python
April 2026 · Certificate ID: ZVO5ZmCKQY-YfG2iC6wfzw
Education
Applied Certificate in Website Design and Development
Saskatchewan Polytechnic
Bachelor of Science
Ahsanullah University of Science & Technology
WES Canadian equivalency: 4-year bachelor