Skip to main content
About · Experience · Philosophy

4+ years of engineering & study, with 1.5+ years shipping production AI.

Junior ML Engineer focused on end-to-end autonomous systems — agents that plan, retrieve, and execute; local LLM inference that scales; and pipelines that hold up in production.

Experience Timeline

From study in 2022 to shipping autonomous AI systems today

  1. Engineering Milestone2025

    AI Hackathon Winner · Performance Excellence · Certified AI Innovator

    Recognized for shipping production-grade AI systems and for cost-eliminating LLM deployment work at Business Automation Ltd.

    • 1st Runner-Up — Business Automation AI Hackathon 2025.
    • Performance Excellence Award 2025 — localized LLM quantization that reduced 100% of API costs.
    • Certified AI Innovator.
    AI EngineeringQuantizationInnovation
  2. Professional RoleOct 2024 — Present

    Machine Learning Engineer (Full Stack AI) — Business Automation Ltd.

    Aggregated 4 years of total experience (study + engineering journey starting in 2022) with over 1.5+ years of dedicated professional industrial experience, strictly focusing on end-to-end autonomous agent architectures over standard LLM API implementations.

    • Integrated complex LLM models into 24+ enterprise web systems across production tenants.
    • Implemented context-aware RAG pipelines and Vector DBs (Pinecone, Weaviate).
    • Orchestrated multi-agent workflows with LangChain / LangGraph and HITL checkpoints.
    • Architected FastAPI / PostgreSQL backends with CI/CD on AWS / Linux / Docker.
    • Performance Excellence 2025 — localized LLM quantization eliminated 100% of API costs.
    PythonFastAPIPostgreSQLLangChainLangGraphPineconeWeaviateAWSDockerLinux
  3. Professional RoleJun 2024 — Oct 2024

    Machine Learning Intern — Business Automation Ltd.

    Statistical modeling and exploratory data analysis for product-facing ML systems. Trained deep learning networks with PyTorch and TensorFlow for downstream GenAI fine-tuning work.

    • Performed statistical modeling and EDA on internal product datasets.
    • Trained and evaluated deep learning networks with PyTorch and TensorFlow.
    • Prepared data pipelines and preprocessing for GenAI fine-tuning experiments.
    • Converted internship into a full-time ML Engineer role in October 2024.
    PythonPyTorchTensorFlowscikit-learnPandasNumPy
  4. Engineering MilestoneJun 2022 — Dec 2024

    Machine Learning Researcher — DIU ML Research Lab

    Long-form research under the DIU ML Research Lab. Built a YOLO-based 9-class traffic violation detection system that reached 98% accuracy — the foundation of the engineering journey.

    • Designed and trained a YOLO-based 9-class traffic violation detection system.
    • Curated and annotated the custom dataset used to train the model.
    • Iterated on accuracy and latency until the system reached 98% accuracy.
    • Thesis: Real-Time Traffic Violation Detection Using YOLO.
    PythonPyTorchYOLOOpenCVLabelImg
  5. Study & Foundation2020 — 2024

    B.Sc. in Computer Science & Engineering — Daffodil International University

    Four-year undergraduate program in Computer Science & Engineering. Thesis focus on real-time computer vision; graduated in 2024.

    • Completed core CS, mathematics, and software engineering coursework.
    • Final-year thesis: Real-Time Traffic Violation Detection Using YOLO.
    • Graduated 2024 with B.Sc. in Computer Science & Engineering.
    CC++JavaPythonData StructuresAlgorithms

Skills Radar

Production-grade across the AI systems stack

Backend & DB

Production APIs, relational engines, and graph stores powering AI services.

  • Python96%
  • FastAPI92%
  • PostgreSQL88%
  • Flask82%
  • DuckDB80%
  • Neo4j78%

Knowledge Retrieval & Vector DB

Grounded RAG systems over vector and graph indexes with semantic + hybrid search.

  • RAG94%
  • Pinecone90%
  • Semantic Search90%
  • Hybrid Retrieval88%
  • Weaviate86%
  • GraphRAG82%
  • Qdrant82%

Agentic AI & Orchestration

Multi-agent systems, function calling, and human-in-the-loop orchestration frameworks.

  • LangChain94%
  • Function Calling92%
  • LangGraph90%
  • Multi-Agent Systems90%
  • HITL84%
  • OpenClaw80%
  • Rasa80%
  • n8n78%

GenAI & Voice AI

LLM adaptation (PEFT, quantization) and real-time voice / TTS / STT pipelines.

  • PEFT (LoRA, QLoRA)86%
  • Quantization (GPTQ, AWQ, GGUF)86%
  • TTS / STT84%
  • LiveKit82%
  • WebRTC80%

LLM Infra

Local GPU inference with KV-cache + runtime buffer pre-allocation, plus API routing.

  • OpenAI / Groq APIs92%
  • vLLM (KV cache + runtime buffer pre-allocation)90%
  • Ollama88%

Vision / OCR

Detection, recognition, and document extraction across print and handwriting.

  • PyTorch90%
  • YOLOv888%
  • OpenCV88%
  • PaddleOCR84%
  • Tesseract78%

Cloud / DevOps

Container orchestration, CI/CD, experiment tracking, and Linux production runtime.

  • Docker90%
  • Ubuntu Linux88%
  • Compose86%
  • Celery86%
  • AWS82%
  • MLflow78%

Engineering Philosophy

What I optimize for when systems hit reality

  • 01End-to-end ownership — from data ingestion and model behavior to deployment, observability, and iteration. Autonomous systems, not isolated notebooks.
  • 02Proof of work over claims: shipped agents, retrievers, and inference engines that run in production, not API-wrapped demos.
  • 03Local-first inference and reproducible deployments. vLLM, Docker, and typed contracts over opaque vendor surfaces.
  • 04Reliability as a feature: typed tool contracts, bounded-cost guardrails, retries, and observability baked in from the first commit.