Skip to content
M. Nobinur — home

AI EngineerDhaka, Bangladesh

M. Nobinur

I build agentic AI and retrieval systems — and the evaluation harnesses that tell you whether they actually work.

Capabilities

What I reach for daily, what I'm comfortable in, and what I'm still learning — each one labelled, so you can tell the difference without asking.

Programming

  • Python, SQLPrimary tool
  • JavaScript / TypeScript, C++Comfortable
  • R, MATLABFamiliar

AI & Agents

  • RAG, LangGraph, LangChainPrimary tool
  • CrewAI, Vercel AI SDK, Mem0Comfortable

ML & Data

  • pandas, NumPy, Evaluation & benchmarkingPrimary tool
  • PyTorchComfortable

Backend

  • FastAPIPrimary tool
  • Django, Next.jsComfortable

Data & Cloud

  • PostgreSQLPrimary tool
  • MySQL, Google Cloud PlatformComfortable
  • DuckDBFamiliar

Tooling

  • Git, Linux / UnixPrimary tool
  • DockerComfortable

Employment

Experience

Two years at Jade & Sterling, across a promotion from software engineering into the AI workflow layer — which I now own end to end.

Jade & Sterling

Aug 2024 — Present · Dhaka, Bangladesh
  1. AI Engineer

    Jul 2025 — Present

    Own the AI workflow layer end to end — retrieval and agent orchestration through to the evaluation harnesses that gate a release.

    • Design and ship production RAG and multi-agent systems with LangChain, LangGraph, CrewAI, and Mem0 — combining retrieval, tool use, orchestration, and persistent memory into reusable services rather than one-off scripts.
    • Built the evaluation layer these systems are judged on: retrieval-quality scoring, response-consistency checks, and deterministic regression suites that catch silent degradation before it reaches users.
    • Expose workflows as versioned FastAPI services with typed request/response contracts and explicit failure handling, so downstream teams integrate against a stable interface instead of a moving target.
    • Containerise and deploy with Docker on Google Cloud Platform, giving every experiment a reproducible path from notebook to production.
  2. Jr. Software Engineer

    Aug 2024 — Jun 2025

    Backend and full-stack delivery across internal services and client-facing product features.

    • Shipped backend services and web features across Python, Django, FastAPI, JavaScript, Next.js, PostgreSQL, and MySQL.
    • Integrated third-party APIs and modelled the database-backed flows behind them, replacing manual reconciliation between internal services and user-facing features with automated data exchange.
    • Debugged and containerised services in Linux environments, and used Git-based review to keep team delivery predictable.

Selected Work

All work →

Researcher

Visual Saliency Evaluation for ASD and Typically Developing Groups

Research · 2025
Benchmarked three saliency models against ASD and TD fixation maps to see whether standard metrics agree on which model predicts which group.
  • PyTorch
  • MATLAB
  • Google Colab
  • NumPy

Researcher

MutOracle-RAG

Research · 2026
A stage-aware evaluation and fault-localization pipeline that attributes RAG failures to retrieval, generation, or the oracle itself.
  • Python
  • PyTorch
  • DuckDB
  • RAG Evaluation

Researcher

Version-Aware Lazy Segment Tree

Research · 2026
Combining range-add lazy propagation with partial persistence, so historical range-sum queries stay correct — built against a brute-force oracle and verified across four platforms.
  • C++17
  • CMake
  • GoogleTest
  • Sanitizers

Founder

sIndeX

Acquired · 2024
An automated SEO platform that replaced recurring manual search-performance analysis with reusable Google API workflows. Acquired in 2024.
  • Python
  • Google APIs
  • SEO Automation

Education

Credentials →
  1. M.S. in Computer Science and Engineering

    North South University · Dhaka, Bangladesh

    2025 — Present

    Graduate coursework chosen around software correctness and experiment design — the two disciplines production AI work keeps asking for.

  2. B.Sc. (Honours) in Statistics

    University of Chittagong · Chittagong, Bangladesh

    2018 — 2024

    Four years of estimation, inference, and the ways data misleads you — the foundation I now apply to reading machine-learning benchmarks for what they actually measure.

  3. Higher Secondary School Certificate (HSC), Science

    Shahid Syed Nazrul Islam College · Mymensingh, Bangladesh

    2015 — 2017

    Science group — the mathematics and physics groundwork that later made both statistics and algorithms approachable rather than intimidating.

When Your Metrics Disagree: Evaluating Saliency Models on ASD and TD Fixation Data

18 Nov 2025 · 7 min
Four standard saliency metrics ranked the same three models differently on ASD and TD fixation maps, and the disagreement turned out to be the finding rather than the noise.
  • Research
  • Computer Vision
  • Evaluation

Shipping RAG That Survives Contact With Real Users

26 Aug 2025 · 8 min
A retrieval-augmented demo and a retrieval-augmented service fail in different ways, and most of the engineering work lives in the gap between them.
  • RAG
  • Engineering
  • Evaluation

What Building sIndeX Taught Me About Handing Over Code

12 Jun 2024 · 6 min
sIndeX started as a script to automate my own repetitive SEO analysis, and the acquisition taught me that undocumented knowledge is a liability with a price attached.
  • Building
  • Career

Contact