Hi, I'm Siam.
I build AI and web applications.
Ex-Grab, Ex-AskTuring AI
Founder of K22B, a fast-shipping web-dev garage
RAG, API endpoints, observability → jack of many trades
MNC-grade engineering rigour, early-stage startup velocity
Likes eating biriyani and reading books
Python · Go · LangChain · LGTM · Dramatiq · Kubernetes · Azure · AWS
01 Experience
- Feb 2025 – May 2026
Full Stack Engineer · AskTuring.ai
California, United States (remote)
- ▹ Led a cross-functional epic that enhanced email ingestion and retrieval in a production RAG architecture.
- ▹ Redesigned the file upload and ingestion pipeline, raising reliability from 70% to 99.95%+ with Dramatiq-based distributed workflows, retries, and reconciler workers.
- ▹ Implemented end-to-end observability with the LGTM stack (Loki, Grafana, Tempo, Mimir) for real-time monitoring and faster incident response.
- Python
- Azure
- Docker
- Dramatiq
- LangChain
- CrewAI
- LGTM
- Aug 2024 – Feb 2025
Full Stack Engineer, Internship · Grab (MyTeksi Sdn Bhd)
Kuala Lumpur, Malaysia
- ▹ Fixed a critical MLflow access control bug, eliminating daily manual password resets for 150+ data scientists and engineers.
- ▹ Built MLOps tools with the AI Infrastructure team, supporting 100+ data scientists across Grab.
- ▹ Deployed OIDC-based authentication with LDAP group access for MLflow, streamlining experiment tracking and service integration.
- ▹ Built endpoints for a central API gateway and feature-serving tool delivering refreshed data to ML pipelines.
- Go
- Python
- Kubernetes
- AWS
- MLflow
- OIDC
02 GitHub Activity
View profile →2,013 contributions in the last year
03 Featured projects
All projects →RAG Ingestion Reliability at AskTuring
Full Stack Engineer · AskTuring.ai · Feb 2025 – May 2026
Redesigned a production RAG file ingestion pipeline from 70% to 99.95%+ success rate using Dramatiq-based distributed workflows with retries and reconciler workers.
- Python
- Dramatiq
- Azure
- Docker
- LangChain
- CrewAI
MLOps Platform Work at Grab
Full Stack Engineer, Internship · MyTeksi Sdn Bhd (Grab) · Aug 2024 – Feb 2025
Fixed a critical MLflow access control bug, deployed OIDC-based auth with LDAP group access, and built endpoints for a central API gateway serving 100+ data scientists.
- Go
- Python
- Kubernetes
- AWS
- MLflow
- OIDC
Narrative Analysis of the July Revolution
CS Researcher · BARTA Lab · May 2025 – Present
Built the largest known dataset of news articles covering Bangladesh's July Revolution and analyzed temporal shifts in media narratives, newspaper stances, and mainstream-vs-social-media coverage.
- Python
- Web Scraping
- Stance Classification
- NLP
Vento: Inventory & Manufacturing Management System
Solo build · May 2026 – Present
A full-stack inventory and manufacturing management system: raw materials, products with recipes, manufacturing runs with history, and a role-aware admin panel. FastAPI + PostgreSQL backend, React + TypeScript frontend.
- Python
- FastAPI
- PostgreSQL
- React
- TypeScript
- Tailwind CSS
Top Words: Trending Bengali Language Analysis
Full-stack build · International Mother Language Institute + BARTA lab collaboration · Aug 2025 – Nov 2025
A full-stack system that extracts and ranks trending Bengali words and phrases from news and social media using N-gram frequency, TF-IDF, and a multi-factor scoring algorithm. Built in collaboration with the International Mother Language Institute and BARTA lab.
- Python
- FastAPI
- SQLAlchemy
- Alembic
- NLP
- React
Opportunity Octopus
Solo build · Apr 2024 – Jun 2024
A full-stack web application centralizing curated opportunities for students and early-career professionals. Built with Django and Tailwind, containerized with Docker, deployed to AWS EC2.
- Python
- Django
- Tailwind CSS
- SQLite
- Docker
- AWS
04 Writing
All posts →-
Which teams got the easy brackets in this World Cup?
I ran the knockout draw through the Elo rating system to find out who got a soft road to the final and who got robbed.
-
How to Implement Facial Recognition with Python
A face-recognition attendance system in Django, end to end. It is easier than you think.
-
Building RICO: Rather Intelligent Conversational Operator
My first shot at building a JARVIS: a voice-controlled assistant in Python.
05 Get in touch
I'm open to AI infrastructure, backend, and full-stack roles. If you're hiring, building something interesting, or just want to compare notes on RAG reliability, I'd love to talk.
Book a 15-min intro call or send an email. Both reach me.