Skip to main content
AI/ML Data Specialist & Startup Founder

Chamika U.

Building high-precision AI dataset evaluation pipelines, scalable full-stack web platforms, and data-driven business intelligence. Founder of SoftNox.

0

Leads Processed via SoftNox

0

AI Annotation Precision

0

Zero-Downtime Records Migrated

AI / ML Expertise

High-Precision
AI Pipelines.

End-to-end dataset engineering for computer vision — from raw image ingestion and CVAT annotation workflows to YOLO model training and Roboflow-managed pipeline quality gates. Consistently delivering above 98% annotation precision at scale.

CVATRoboflowYOLOv8LabelStudioPythonPyTorch
CVAT bounding box annotation — multi-class object detection at 98% precision
Object DetectionCVAT · YOLOv8 · 500K+ annotations
Dataset evaluation dashboard — confusion matrix and precision/recall metrics
Dataset EvaluationmAP@0.5: 0.961 · Precision: 98.2%
YOLO model training pipeline — loss curve and mAP progression over 200 epochs
Model Training PipelineEpoch 142/200 · loss: 0.0312
Applied Architecture
01 — Data Preparation & Evaluation

Structuring data for
computer vision.

Hands-on experience evaluating and annotating complex AI datasets for computer vision pipelines. Working within CVAT and Roboflow workflows to produce precise, schema-consistent training data — bounding boxes, segmentation masks, and multi-class label hierarchies — directly consumed by YOLOv8 training runs. Emphasis on annotation accuracy, class balance, and dataset versioning hygiene.

View annotation projects
02 — Backend Architecture

Executing seamless
data migrations.

Designed and executed the extraction, transformation, and structured migration of static JSON record sets into live PostgreSQL databases — normalising schema, enforcing relational integrity constraints, and validating row-level fidelity post-migration. Built with repeatability and zero data loss as the non-negotiable primary constraint.

View migration scripts
Partner Feedback

“The ETL pipeline Chamika architected cut our data ingestion latency by 60%. He thinks architecturally about every layer — not just the deliverable.”

Marcus T.

Head of Data, Synapse Analytics

“Our computer vision dataset went from noisy ground truth to publication-ready in weeks. Every bounding box validated against strict protocols.”

Priya R.

ML Research Lead, Visionix AI

“Schema design was flawless. The migration ran without a single hour of downtime across 50 million production records. Remarkable precision.”

James K.

VP Engineering, Meridian Data

“The annotation quality exceeded every benchmark we set. Chamika’s QA process caught errors our own team had missed for months.”

Selin A.

Data Science Director, Lumio Labs

“The ETL pipeline Chamika architected cut our data ingestion latency by 60%. He thinks architecturally about every layer — not just the deliverable.”

Marcus T.

Head of Data, Synapse Analytics

“Our computer vision dataset went from noisy ground truth to publication-ready in weeks. Every bounding box validated against strict protocols.”

Priya R.

ML Research Lead, Visionix AI

“Schema design was flawless. The migration ran without a single hour of downtime across 50 million production records. Remarkable precision.”

James K.

VP Engineering, Meridian Data

“The annotation quality exceeded every benchmark we set. Chamika’s QA process caught errors our own team had missed for months.”

Selin A.

Data Science Director, Lumio Labs

“Fast, precise, and extraordinarily professional. The dashboard handles half a million records without a stutter — exactly what high-scale operations demand.”

Daniel O.

CTO, DataPulse Systems

“The NLP annotation pipeline reduced our model fine-tuning cycles from weeks to days. Ground-truth data at a quality I didn't think was achievable externally.”

Aisha M.

Head of NLP, Korvex AI

“Chamika delivered a full-stack auth system in half the time we estimated. Clean code, zero security gaps, and exceptionally documented.”

Ryan W.

Founder, Stackframe

“Orchestration logic across four data sources — connected, validated, and live in under two weeks. An engineer who actually understands data at scale.”

Nadia F.

Analytics Lead, Orbis BI

“Fast, precise, and extraordinarily professional. The dashboard handles half a million records without a stutter — exactly what high-scale operations demand.”

Daniel O.

CTO, DataPulse Systems

“The NLP annotation pipeline reduced our model fine-tuning cycles from weeks to days. Ground-truth data at a quality I didn't think was achievable externally.”

Aisha M.

Head of NLP, Korvex AI

“Chamika delivered a full-stack auth system in half the time we estimated. Clean code, zero security gaps, and exceptionally documented.”

Ryan W.

Founder, Stackframe

“Orchestration logic across four data sources — connected, validated, and live in under two weeks. An engineer who actually understands data at scale.”

Nadia F.

Analytics Lead, Orbis BI

What's Next

Let's build something
worth scaling.

Open to collaborative projects, contract work, and opportunities at the intersection of data engineering, AI pipelines, and scalable full-stack architecture. Based in Colombo, LK — available globally.

Get in Touch