ABCDullahh
Initializing
Faizal Lutfi
If the right tool exists, build with it.
If it doesn't — understand the problem, then build the tool.
About
AI Software Engineer with 4+ years of professional experience — spanning construction data analytics to full-stack AI systems. My career path is unconventional but deliberate: every role sharpened a different edge, and the main focus has always been IT and AI.
I work end-to-end — from problem definition to deployment — building reliable solutions across LLM orchestration, applied ML, and backend systems that drive measurable business outcomes.
Currently at a Mitsubishi Group IT consulting company, shipping enterprise AI tools — RAG pipelines, automated transcription systems, and AI-powered training platforms. Previously designed highlight detection AI at a gaming media-tech startup.
Technology evolves in real time — AI alone has reshaped the landscape in just three years.
Stay current, adapt fast, or get left behind.
0+
Years Experience
0+
Projects Shipped
Experience
AI Software Engineer
NOWPT Berlian Sistem Informasi
Jun 2025 — Present
Building AI-driven enterprise solutions at a Mitsubishi Corporation Group IT consulting company. Working on LLM + RAG systems, internal tooling, and production-grade AI pipelines for business automation.
AI & Data Optimization
Eklipse.gg
Feb 2025 — Apr 2025
Designed AI architecture for automated highlight detection in gaming video content at an AI video editing media tech company. Focused on training workflows, data pipelines, and LLM-assisted video analysis.
Data Analyst & Python Developer
Berkah Prima Perkasa
Feb 2024 — Feb 2025
Performed data analysis and built Python tooling for a construction materials supplier. Focused on inventory optimization, logistics analysis, and data-driven decision making across cross-functional teams.
Web Developer & Front-end
Tegar Ponsel
Jan 2023 — May 2023
Designed and developed a responsive e-commerce interface for a mobile phone sales and distribution company. Focused on front-end development and API integration.
Systems I've Built
Production enterprise systems handling real users at scale — designed, built, and deployed by me.
EIRA
PRODUCTIONEnterprise Intelligent Response Assistant
Problem
Engineers and business staff across a Mitsubishi Group IT company had no centralized AI tool — relying on fragmented generic chatbots with no access to internal documentation, code context, or company-specific workflows. Knowledge was siloed and repeated questions wasted hours daily.
Approach
Designed and built a multi-LLM enterprise assistant with 7+ custom feature modules: intelligent code analysis, automated document generation, workflow automation, MoM transcription, RAG-powered Q&A, data visualization, and report generation. Powered by Gemini, OpenAI, and Azure AI Foundry with dynamic model routing. Secured via multitenant MSAL authentication with Azure AD Entra enterprise application — enabling SSO across Mitsubishi Group subsidiaries with role-based access per tenant.
Impact
Adopted by 1,000+ employees weekly across engineering, HR, and business divisions. Sub-2-second streaming responses. Unified 7+ fragmented tools into one platform. Multitenant SSO across subsidiaries via Azure AD Entra. Custom features tailored per department — code review for engineers, document drafting for business, meeting notes for managers.
1000+
users/week
Multi
LLM routing
7+
custom features
System Architecture
MoM PIPELINE
PRODUCTIONAutomated Minutes of Meeting
Problem
Recording and transcribing lengthy meetings (1-3+ hours) was a manual, error-prone process. Most STT APIs like Whisper enforce a strict 20-25MB upload limit — simply splitting audio at arbitrary points causes hallucination, cut-off sentences, and lost context. Additionally, raw WAV files from recorders are massive and need format conversion before processing.
Approach
Engineered an intelligent audio preprocessing pipeline: WAV/MP3/M4A files are first normalized via FFmpeg, then the waveform is analyzed to detect silence segments — the quietest points between speakers or natural pauses. Audio is split precisely at these silence boundaries, ensuring each chunk contains complete sentences. A pooling queue system manages concurrent transcription requests, preventing API rate limits and handling 1,000+ simultaneous uploads. Each chunk is transcribed via Whisper, then an LLM assembles the full transcript into a standardized MoM format.
Impact
Handles audio uploads up to 200MB+ (including raw WAV) seamlessly — 10x the API limit. Pooling system supports 1,000+ concurrent meeting transcriptions with queue-based load balancing. Clean, structured MoM documents generated in minutes. Integrated directly into EIRA as a custom feature module.
200MB+
audio support
1000+
concurrent
10x
API limit bypass
System Architecture
AI TRAINER
PRODUCTIONSales Training Intelligence
Problem
Onboarding and upskilling sales teams across Mitsubishi's nationwide distribution network was inconsistent, expensive, and unscalable — manual roleplay sessions delivered limited feedback with no objective measurement, making it impossible to track improvement or standardize training quality.
Approach
Architected an AI-driven training platform where Azure AI Foundry LLM agents simulate realistic customer personas with ElevenLabs voice synthesis for natural conversation. Each session is recorded, transcribed, and automatically evaluated by an LLM scoring engine that generates structured coaching feedback with actionable improvement points.
Impact
100 concurrent trainees supported simultaneously in real-time sessions. AI-based scoring replaces subjective human evaluation — consistent, objective, and instant. Designed for national rollout across all Mitsubishi sales divisions with multi-scenario support for different product lines.
100
concurrent
Voice
AI synthesis
National
scale rollout
System Architecture
Projects
13 projects — hover to explore
Education & Certifications
Diponegoro University
2020 — 2024
Computer Engineering
Bachelor’s degree in Computer Engineering from Diponegoro University, specializing in coding, database management, and web application development. My academic experience has provided me with strong technical skills and a deep understanding of how to design, develop, and manage web-based systems. I am passionate about creating efficient and scalable solutions, always eager to apply my knowledge in real-world projects
Purwadhika Digital Technology School
2024 — 2024
Data Science & Machine Learning
Completed an intensive program focused on data analysis, SQL, and machine learning. Developed and deployed predictive models to address real-world business challenges, enhancing operational efficiency and supporting strategic decision-making through data-driven solutions.
Certifications
Tech Stack
Contact
Let's Build
Something Great
Jakarta, Indonesia
© 2026 Faizal Lutfi