Available for US Remote Roles & Engineering Architecture

Senior Full Stack & Systems Engineer
MedTech & Distributed Systems

Senior Full Stack & Systems Engineer with 5+ years of production experience building medical imaging pipelines, low-latency streaming architectures, and fault-tolerant distributed platforms. Specialized in taking complex Software as a Medical Device (SaMD) through technical verification and FDA 510(k) clearances (NeuroICH, MammoSightAI, CXRDetectAI, RadioView.AI). Track record managing 29M+ imaging instances, cutting critical alert latency by 98% (24.5 min to 24s), and scaling cloud systems across Python/FastAPI, Node.js, Next.js, Flutter, and PostgreSQL/Supabase.

Explore FDA ClearancesView Projects
architect.config.ts
const architect = {
role: "Senior Full Stack & Systems Engineer | MedTech & Distributed Systems",
location: "Dhaka, Bangladesh (US Remote)",
scale: "29M+ Instances",
activeClearances: ["K241719", "K252954", "K253751"],
coreStack: ["Python", "FastAPI", "React", "Next.js"],
triageVelocity: "98% Drop (Emergency notification latency cut from 24.5 min to 24s)"
};
STATUS: 200 OKSTRICT SCHEMA
5+ Yrs
Industry Experience
Distributed systems & health tech engineering
29M+ Instances
Platform Scale
30 microservices · 1,000+ clinicians · 99.99% uptime
98% Drop
Triage Acceleration
Emergency notification latency cut from 24.5 min to 24s
40+ Stars
Open Source Reach
Systems utilities, CLI tools & IPC wrappers

FDA 510(k) Clearances & Clinical AI

SaMD Certified

SaveLife.AI Suite™ & RadioView.AI™ Cloud PACS Verification & Validation

510(k)Public Regulatory Records
NeuroICH™K241719
Non-Contrast Head CT
97.5% Sensitivity · <30s Alert

AI-driven intracranial hemorrhage detection and emergency triage on non-contrast head CT scans, validated across Siemens and Philips hardware with PACS worklist priority orchestration.

MammoSightAI™K252954
Screening & Diagnostic Mammography
Lesion Localization & Density Profiling

Clinical AI module for digital mammography, delivering automated lesion identification, multi-view spatial correlation, and breast tissue density assessment markers.

CXRDetectAI™K253751
Digital Chest Radiography (CXR)
Multi-Pathology Emergency Triage

Deep learning triage engine detecting pneumothorax, pleural effusion, and lung consolidation with automated urgent flagging to radiologists.

RadioView.AI™ Cloud PACSFDA Cleared
Diagnostic Radiology Workstation
24s Critical Alert · 99.1% Abnormal Sensitivity

FDA-cleared cloud PACS with real-time AI-Findings triage and bidirectional DICOMweb integration; slashes emergency radiologist critical findings notification from 24.5 min to 24s.

DenseViewAI™ & NeuroLVO™510(k) Under Review
Breast Density & Stroke CTA
Clinical Multi-Reader Validations

Automated breast density scoring and acute ischemic stroke large vessel occlusion (LVO) triage undergoing FDA pre-market clearance submissions.

Co-engineered SaMD software verification & validation (V&V) harnesses, clinical data pipelines, and design control traceability.
IEC 62304 & 21 CFR Part 11

Distributed Architecture Deep-Dive

RadioView.AI / Cloud PACS

End-to-End Clinical Ingestion, Automated Anonymization, Async AI Inference & Diagnostic Streaming

STAGE 01·

DICOM Ingestion & Gateway

High-throughput ingest of clinical CT/MRI/X-ray multi-slice series via WADO-RS & DICOM-TLS.

Performance SLA Target<200ms per series ingest · 99.99% availability
Architectural Decision & Trade-off
Chosen ApproachChunked multipart streaming over HTTP/2 with DICOM-TLS
Alternative RejectedMonolithic multipart archive uploads

Engineering Rationale: Prevents memory ballooning on 2,000+ slice CT scans and enables resumable transfers across hospital edge nodes over lossy institutional WANs.

Production Metrics & Verification
  • 29M+ DICOM instances ingested
  • 30 distributed microservices
Technologies
FastAPIDICOMwebTLS 1.3Docker

Professional Engineering & Leadership

5+ Years Production Experience · Distributed US & Global Roles

Team Lead & Senior Full Stack Engineer

Full-Time
SaveLife.AI

US medical imaging AI startup delivering real-time clinical triage and cloud PACS to 1,000+ clinicians.

US-Based (Remote)Jan 2023 – Present
4 FDA Clearances29M+ DICOM Instances30 Microservices24s Triage Time
  • Cloud PACS Workstation Architecture: Architected the diagnostic imaging viewer and clinical triage dispatch for RadioView.AI™ across a 30-microservice architecture managing 29M+ DICOM instances and 1M+ clinical reports with 99.99% operational uptime.
  • Low-Latency Notification Dispatch: Engineered real-time event conduits (SSE/WebSockets) lowering emergency radiologist critical findings notification from 24.5 minutes to 24 seconds (98% reduction) with 99.1% abnormal detection sensitivity.
  • FDA 510(k) Software Verification: Built automated end-to-end test suites and clinical validation data pipelines for NeuroICH™ (K241719), MammoSightAI™ (K252954), CXRDetectAI™ (K253751), and RadioView.AI™, supporting active submissions for DenseViewAI™ and NeuroLVO™.
  • Regulatory Controls & Security Governance: Authored software lifecycle documentation (IEC 62304) and automated 21 CFR Part 11 audit trails; implemented zero-trust access controls and encrypted DICOM-TLS pipelines to sustain HITRUST CSF and HIPAA compliance.
  • Engineering Leadership & Mentorship: Lead day-to-day sprint delivery for a distributed squad of 6 engineers; conducted code reviews, standardized CI/CD pipelines, and bridged product requirements between regulatory affairs and backend engineers.
Tech Stack:PythonFastAPIReactNext.jsFlutterSupabaseDICOMwebDocker21 CFR Part 11WebSockets

Open Source & Engineering Projects

Production-Grade Tooling, Medical Imaging CLIs & AI Apps

adb_ui

41
Creator & Maintainer

Cross-platform desktop GUI for Android Debug Bridge with native IPC wrappers.

Cross-platform desktop systems GUI for Android Debug Bridge; features native IPC subprocess wrappers, asynchronous command queueing, wireless Wi-Fi pairing protocols, and embedded scrcpy screen mirroring.

Problem: Managing multiple Android test devices via terminal ADB commands is cumbersome and slow for mobile QA and developer workflows.
Trade-off: Native asynchronous IPC subprocess wrapping (C++/Dart) vs. spawning heavyweight external shells; chosen for sub-5ms command dispatch and robust error capture.
Outcome: 40+ GitHub stars, seamless Wi-Fi ADB pairing, APK drag-and-drop installation, and integrated scrcpy screen mirroring without GUI thread starvation.
FlutterDartAndroid Debug Bridge+3
Code

AI-Chat

Creator

Production conversational platform with multi-provider routing and SSE streaming.

Production conversational platform featuring multi-provider LLM dynamic routing, low-latency Server-Sent Events (SSE) streaming protocols, resilient markdown/LaTeX parsers, and edge-persisted session memory.

Problem: Multi-provider conversational LLM clients suffer from UI lag, dropped connections during token streaming, and inconsistent token usage tracking across vendors.
Trade-off: Server-Sent Events (SSE) with Edge token buffer vs. bidirectional WebSockets; chosen for native HTTP/2 multiplexing, automatic reconnection, and zero stateful connection server overhead.
Outcome: Sub-100ms time-to-first-token (TTFT), resilient markdown/LaTeX math rendering, and multi-model failover routing across OpenAI, Claude, and local Ollama instances.
Next.js 16TypeScriptReact 19+3
Code

dicom-scripts

Creator & Maintainer

High-performance Python imaging utility for de-identification and UID remapping.

High-performance imaging metadata extraction and de-identification engine; performs in-place binary tag manipulation, memory-efficient streaming anonymization, and deterministic UID generation for clinical datasets.

Problem: Clinical DICOM studies frequently contain patient Protected Health Information (PHI) burned into binary tags, causing compliance violations when migrating data to research environments.
Trade-off: In-place binary tag mutation using pydicom streaming vs. full file uncompressed re-encoding; chosen to minimize memory footprint on multi-gigabyte CT studies.
Outcome: Deterministic UID remapping and memory-efficient de-identification processing 500+ DICOM instances/sec with zero memory leaks.
PythonpydicomDICOM+3
Code

GameHunter

Creator

Desktop hardware profiler and game discovery suite interfacing with OS hardware counters.

Desktop discovery platform and local PC hardware profiler; directly queries native OS hardware counters to evaluate CPU/GPU bottlenecking and match real-time system specifications against demanding catalog titles.

Problem: Users frequently experience game crashes or severe bottlenecking from mismatched GPU/CPU hardware configurations.
Trade-off: Querying native Windows WMI and DirectX performance counters directly vs. generic browser-based system queries; chosen for accurate hardware telemetry.
Outcome: Real-time CPU/GPU bottleneck calculations and automated specification matching across 10,000+ games.
FlutterPythonFastAPI+2
Code

audio_switcher

Creator

Zero-overhead Windows systems utility interfacing with Core Audio COM APIs.

Zero-overhead Windows systems utility interfacing directly with Windows Core Audio COM interfaces for sub-5ms device switching via global hotkey macros without GUI thread blocking.

Problem: Toggling Windows audio playback endpoints normally requires navigating slow multi-click OS control panels, interrupting games and recording sessions.
Trade-off: Direct Windows Core Audio COM interface invocation via lightweight C++/PowerShell wrappers vs. resident background daemon; chosen for 0% idle CPU and zero memory residency.
Outcome: Sub-5ms instant audio endpoint switching bound to global macro hotkeys with zero background overhead.
PowerShellBatchWindows Core Audio COM+2
Code

Red Door Discord Bot

Creator

High-concurrency containerized Discord bot leveraging Dart event-loop concurrency.

High-concurrency containerized Discord bot leveraging Dart's event-loop concurrency model; features sharded WebSocket gateway connections, interactive components, and zero-downtime Docker orchestration.

Problem: High-volume Discord bots frequently drop incoming message gateway events during concurrent spikes and thread pool exhaustion.
Trade-off: Single-threaded event-loop concurrency (Dart nyxx) with sharded WebSocket gateways vs. multi-threaded thread-per-message; chosen for predictable memory usage and lock-free async streams.
Outcome: 99.9% gateway connection uptime with automated reconnect backoff and containerized zero-downtime Docker deployments.
Dartnyxx frameworkDocker+2
Code

Core Competencies & Technology Matrix

Regulatory Standards, Distributed Cloud & Medical AI Pipelines

Interactive Matrix

Distributed Systems & Backend

High-throughput microservices, event-driven architectures, and resilient cloud databases.

8 Capabilities
Python (FastAPI)
Node.js & Express.js
Supabase & PostgreSQL
Microservices Architecture
Event-Driven & SSE / WebSockets
RESTful APIs & GraphQL
Database Query Optimization
Zero-Trust Service Mesh

Governance, Security & Enterprise Standards

21 CFR Part 11 · HIPAA · HITRUST CSF · IEC 62304

21 CFR Part 11 Audit Trails
Electronic Records, Cryptographic Audit Logs & Signature Verification
Audit Trace
HIPAA & HITRUST CSF
End-to-End PHI Encryption, DICOM-TLS, Zero-Trust Access Control
HITRUST / HIPAA
FDA 510(k) Pre-Market Clearances
NeuroICH (K241719), MammoSightAI (K252954), CXRDetectAI (K253751), RadioView.AI
FDA 510(k)
IEC 62304 / ISO 13485 Principles
Medical Device Software Life Cycle & Design Controls
SaMD Standards

Systems Architecture & Engineering Impact

Low-Level OS Wrappers, IPC & Distributed Reliability

29M+ DICOM Instances & 1M+ Clinical Records99.99% Uptime

Distributed 30-microservice architecture across edge and cloud PACS nodes serving 1,000+ active clinicians.

OS Wrappers & Native IPC SubprocessesLow-Level

Windows Core Audio COM interfaces (<5ms switching), Android Debug Bridge IPC integration, and pydicom binary manipulation.

Demonstrated track record spanning 5+ years of continuous high-impact production engineering delivery.