Description

My journey as a Software Developer, enriched with a BS in Electronics and Telecommunication Engineering, has been marked by a commitment to innovation and the practical application of cutting-edge technologies. As a Machine Learning Engineer allowed me to spearhead the development of no-code AutoML solutions and optimize machine learning models, enhancing performance and user experience.

My work involved receiving datasets, training models through a series of unique steps, and deploying these models directly for client use. This hands-on experience in dataset preparation and model optimization has honed my skills in practical machine learning applications.

With a passion for the Industrial Internet of Things (IoT), edge computing, and security, I am constantly exploring how these fields can intersect with machine learning to create smarter, more secure systems. My expertise in crafting intelligent systems is complemented by a collaborative approach to problem-solving and a commitment to continuous learning.

My experience in back-end development, computer vision, and Python, I possess a unique ability to integrate diverse technologies and methodologies. I am passionate about IoT, cloud computing, information security, and data science, always seeking innovative ideas and striving to push the boundaries of what is possible.

With over three years of freelance software development experience(parttime) and more than two years as a remote ML Engineer/Python Developer(fulltime) at a startup producing no-code AutoML solutions, I have been actively involved in research and development projects in AI/ML/MLOps. My ability to work effectively as part of a team, coupled with a unique perspective on solving complex problems, allows me to make significant and impactful contributions.

Languages

English
Native or bilingual

Workplace preferences

Remote only

Primarily works remotely

Freelance
Software Developer - ML Engineer
January 2024 - Today (2 years and 5 months)
• Engineered feature pipelines with sentiment-subjectivity signals and finetuned classifiers to separate human vs. LLM text (Kaggle: LLM Detect AI Generated Text) Modeled human preference patterns and learned efficient LLM serving with vLLM (continuous batching, paged attention) to improve inference efficiency and judgment quality (Kaggle: LMSYS Arena) Produced high signal annotations for large scale preference-safety evaluation that informed reward model tuning in RLHF workflows: Role focused on consistent judgment quality (Scale AI - LLM Evaluation)
• Designed a lean CTR pipeline with attention based modeling and latency aware inference: Iterated architecture, finetuning to pursue AUC gains and deployability (Custom FINT CTR)
• Built an end to end RAG prototype with vector retrieval, prompt routing and lightweight evaluation, demonstrating reliable grounding and fast iteration (LLM - RAG Application)
• Implemented a real time camera streaming pipeline with pluggable preprocessing and model hooks for rapid computer vision prototyping
• Packaged a fast, training free neural style transfer workflow using develop a Comfy, simplifying stylized image generation scripts
• Created a pose analysis sandbox over human keypoints to explore ergonomics and activity cues: Delivered a practical PoC(pose-thing) Prototyped a lightweight VLM/Gemma coaching workflow to streamline prompting, FT hooks, and evaluation loops for rapid experiments (gemma3n-workbuddy)
• Built a scheduling/coordination interface, capturing constraints and improving execution visibility (Planning & Organization Tool)
• Researched deep learning pipelines for 5G traffic analysis and threat detection, exploring fDNN trade-offs to balance accuracy and speed (5G Cyber Threat Detection)

Computer Vision / Image Processing LLM TensorFlow
Datamarkin
ML Engineer
June 2021 - July 2023 (2 years and 1 month)
• Built an end-to-end no-code CV/AutoML platform that reduced months-long, human-driven workflows to minutes with one-click deploy, enabling bulk processing of millions of samples in under 1 minute and cutting time-to-production by >90%
• Implemented platform-wide model optimization (partial freezing, selective fine-tuning, LR/optimizer tuning, ONNX, TFLite, HPO, transfer learning) to accelerate training ~5x, reduce GPU footprint ~60%, and improve production stability and throughput
• Designed and implemented low-latency on-prem inference with a modular, scalable, high-throughput deployment stack, hosting 1,000+ models per GPU with ~3 s cold start and sustaining >100 req/s per GPU under load
• Delivered production-ready modules for object detection, keypoint detection, and segmentation within no-code workflows; expanded the platform from classification to these tasks with modular, reusable AutoML-aligned components, and shipped optimized models with task-specific data/augmentation for seamless deployment into the serving stack
• Built a cost-efficient, privacy- and security-focused multi-tenant platform with integrated IAM, serving thousands of concurrent users and offering dedicated, isolated servers for enterprise tenants
• Supported academic partners on applied CV projects with no-code AutoML tooling and deployment; partner institutions publicly shared project outcomes and publications
Model Optimization Machine Learning Theory OpenCV Computer Vision Pytorch