SOFTWARE ENGINEER & SYSTEMS ARCHITECT

Ho Cong Toan

Full-stack developer building high-performance backends and intelligent automation systems.

Software Engineering student at PTIT. Building scalable infrastructure and optimized distributed systems.

“Simplicity is the soul of efficiency.”

— Austin Freeman

Education

PTIT · 4th Year

Focus

Software Engineering

Exploring

Microservices & DevOps

Location

Ho Chi Minh City, VN

02

Tech Stack

Technologies I use to design, build, and deploy systems — from low-level concurrency to cloud-native infrastructure.

Languages

Java

Go

Python

TypeScript C++

C++

Backend

Spring Boot

Spring Cloud

Echo (Go)

Flask

Next.js

Node.js

Data & ML

PostgreSQL MySQL

MySQL

Redis

Milvus

TensorFlow

HuggingFace ONNX

ONNX

Infrastructure

Docker

Kafka

Cloudflare AWS

AWS

Nginx

Git

03

Systems & Architecture

What I've designed and built — distributed backends, ML pipelines, cloud infrastructure, and security systems.

BotCV — Job Recruitment Platform

5 SERVICES

Polyglot microservice architecture spanning 3 languages. AI-powered recommendations via collaborative filtering + content-based vector similarity. Event-driven sync via Kafka.

Service Layer

Job API

Spring Boot 3.5 · JPA · FlywayCore REST API — job postings, employers, candidates

Auth Service

Echo v4 · JWT · RSA · KafkaOAuth2, RSA JWT tokens, refresh rotation

User Service

Echo v4 · pgxProfile management, preferences

Resource Service

Echo v4 · CloudinaryCV upload, file storage, NER extraction

AI / ML Layer

Search & Recommendation

Flask · PyMilvus · BGE-M3 · ALSHybrid search (dense + sparse vectors) and 2-stage recommendation engine

Infrastructure

PostgreSQL 18Redis 7.4Milvus 2.4KafkaDocker Compose

▶ Hybrid vector search (dense + BM25 sparse)▶ ALS collaborative filtering + cold-start fallback▶ Event-driven embedding sync via Kafka

DTLN — Real-Time Speech Enhancement

~2M PARAMS

Dual-Signal Transformation LSTM Network for noise suppression, trained on Vietnamese speech. Deployed as a real-time WebSocket microservice for IoT audio processing (<50ms latency).

Dual-Core Pipeline

1STFT Domain

Noisy audio → STFT → magnitude → 2×LSTM(128) → sigmoid mask → masked magnitude → iFFT

2Learned Transform Domain

Conv1D encoder → InstantLayerNorm → 2×LSTM(128) → sigmoid mask → Conv1D decoder → overlap-add

Clean Audio Output → WebSocket Stream

16kHz sample rate512 block / 128 hopNeg-SNR lossAdam + clip(3.0)

Evaluation Metrics

Signal-to-Noise Ratio

14.57dB

10.08 dB+4.49 dB

SI-SDR

14.77dB

9.81 dB+4.97 dB

PESQ

2.01

1.48+0.53

STOI

0.898

0.854+0.044

04

Get In Touch

Always interested in discussing distributed systems, ML engineering, or potential collaboration.

hctoan.dev@gmail.com LinkedIn GitHub