Oshan Mudannayake

Home Research Publications Experience Projects Open Source Certifications Achievements Workshops Talks Volunteering People Articles CV About

Experience

Research Experience

  • Community Researcher

    Cohere For AI · Sri Lanka · Advised by Sara Hooker · Mar. 2023 - Apr. 2024
    • Built the web platform used to collect human annotations and contributed data for the Aya multilingual instruction-tuning dataset; published at ACL 2024.

  • Research Assistant (Contract)

    University of Colombo School of Computing · Colombo · Advised by Dr. Noel Fernando · May 2021 - Dec 2024
    • Applied classical and deep time-series models to forecast long-horizon municipal solid-waste generation in Sri Lankan cities; published in IEEE Access.

  • Researcher (Contract)

    Sustainable Computing Research (SCoRe) Lab · Colombo · Advised by Dr. Mohamed Nabeel · Aug. 2020 - Dec. 2021
    • Developed an ML-based detector for malicious internet domains using lexical, registration, and DNS-resolution features.

  • Research Intern

    University of Colombo School of Computing · Colombo · Advised by Dr. Nalin Ranasinghe · Jul. 2018 - Jan. 2019
    • Prototyped a privacy-preserving training layer over TensorFlow enabling model fitting on homomorphically encrypted tensors across a distributed worker network.
    • Presented the work as a poster at the Sri Lanka Association for the Advancement of Science (SLAAS) Exhibition, Dec. 2018.

Professional Experience

  • Lead ML Engineer

    ParadigmAI · Colombo · Aug 2023 - Present
    • Built an LLM-backed platform that predicts a respondent's survey answers from their prior response history.
    • Designed the translation backend for a low-code system compiling graph-structured workflows into executable code.
    • Engineered a multi-transport LLM gateway (REST/gRPC) with key rotation, vendor fallback (OpenAI/Anthropic), and request-level tracing for production traffic.
    • Implemented a Celery-on-SQS transcript evaluation service with Pydantic-enforced response schemas, streamed scoring via an internal OpenAI proxy, and Postgres-backed result sync with automated feedback generation.
    • Built LLMOps tooling: an async Celery/SQS evaluation service and a Streamlit analytics suite surfacing cost, latency, and health metrics across LLM workloads.
    • Architected an AI-powered code evaluation platform ingesting GitHub repositories for automated review.

  • Senior Data Science Engineer

    WSO2 · Colombo · Nov 2021 - Aug 2023
    • Trained and deployed classification, regression, and clustering models driving product adoption and user-engagement metrics.
    • Built internal analytics dashboards exposing model performance, A/B test outcomes, and business KPIs.
    • Standardized team practice for experiment tracking, model versioning, and reproducible evaluation pipelines.

  • Data Science Engineer

    WSO2 · Colombo · Apr 2021 - Oct 2021
    • Analyzed product-usage telemetry and built predictive models to surface adoption trends; scoped ML opportunities with product and engineering stakeholders.
    • Scoped ML opportunities with product and engineering stakeholders and presented findings to technical and non-technical audiences.

  • Software Engineer - IAM Team

    WSO2 · Colombo · Mar 2020 - Mar 2021
    • Shipped features into WSO2 Identity Server on the product-development track and served on the patch-services team triaging and backporting fixes across supported customer releases.
    • Served on the patch (services) team triaging and backporting security and functional fixes across supported customer releases.

  • Student Developer

    CERN-HSF - Google Summer of Code · Remote · May. 2018 - Aug. 2018
    • Contributed to Phoenix, the next-gen Vue.js web UI for ownCloud and CERNBox (CERN's cloud-storage platform serving 13K+ users over 2.5 PB), with offline-capable browser operation via service-worker caching and a Go-based local WebDAV bridge.
    • Added offline-capable browser operation: service-worker metadata caching, offline-mode detection with reconciliation on reconnect, and a Go-based local WebDAV bridge integrating the desktop sync client to redirect file I/O while disconnected.

  • Software Developer

    Language Matters · Sri Lanka · Advised by Dr. Leonie Solomons · Jan. 2018 - Mar. 2020
    • Developed TessTA, a production OCR system for Sinhala: fine-tuned the Tesseract engine on a hand-curated Sinhala corpus and diagnosed and resolved script-specific failure modes (character decomposition, ligature segmentation, diacritic attachment) to lift recognition accuracy on low-resource Sinhala text.

  • Software Developer

    Dreamteam IT Solutions · Sri Lanka · Advised by Dr. Gihan Kuruppu · Feb. 2016 - Jan. 2017
    • Built an instrumented Smart Glove prototype (FSR-400 pressure array + Arduino) capturing hand pressure distribution and acceleration to model cricket batting grip dynamics.
    • Deployed an embedded OpenCV pipeline on Raspberry Pi detecting knots in yarn threads on industrial glove-weaving machines, eliminating hundreds of needle breakages per day and reducing manual inspection.
    • Built a conveyor-line image-processing system identifying common textile-glove defects in-flow and triggering automated reject handling without operator intervention.

  • Course Instructor

    Dreamteam IT Solutions · Sri Lanka · Jan. 2016 - Jun. 2016
    • Delivered Java preparation sessions for the Oracle Certified Java Programmer (OCJP, Java SE 6) certification, covering the full certification syllabus.
    • Taught an introductory robotics course centered on line-following robots, covering sensor calibration, simple control loops, and microcontroller programming.