Open to AI and product roles

Four years putting models into production — from a research lab in Moscow to a startup in New York.

See the rolesDownload the CV
Portrait of Thomas Gabriel Chung.
+75%response accuracyGdeRadost
1,000+usersGT Augment

Experience

Where the models were built.Research and production.

From a research lab in Moscow to a startup in New York. Open any card for the detail.

7 roles, newest first

  1. 4 mos
    • Growth & Partnerships
    • AI & Product
    Jun 2025 — Sep 2025

    AI B2B Sales

    Key Group · Moscow, Russia

    B2B acquisition using AI tooling — building and optimising outbound campaigns for travel, banking and airline clients.

    • −30%time on routine tasks
    • +40%prospect base
    • 5+AI tools adopted
    • Sourced and won B2B clients using Apollo and LinkedIn Sales Navigator.
    • Launched and optimised email and LinkedIn sequences through Apollo and Expandi.
    • Automated workflows with Zapier, Amplemarket and Clay, plus LLMs for data processing.
    • Analysed campaign results and adapted strategy on the data.
    • Sales automation
    • Apollo
    • Expandi
    • Clay
    • Zapier
  2. 5 mos
    • AI & Product
    Mar 2025 — Jul 2025

    AI Specialist

    QuestMeUp · Florida, USA

    Led AI development for a personalised assessment platform, designing and fine-tuning the models behind its adaptive user journeys.

    • Designed and fine-tuned AI models for personalised assessments and adaptive user journeys.
    • Translated technical workflows into clear documentation for a non-specialist team.
    • Shaped the platform vision and AI roadmap so development stayed aligned to user goals.
    • LLM fine-tuning
    • Product
    • Technical writing
  3. 4 mos
    • AI & Product
    Nov 2024 — Feb 2025

    AI Specialist

    GdeRadost · Moscow, Russia

    Improved model response quality on a production assistant through prompt engineering and fine-tuning.

    • +75%response accuracy
    • Raised AI model response accuracy by 75% through prompt engineering and fine-tuning.
    • Worked with engineers and analysts to ship scalable AI solutions across departments.
    • Drafted internal model evaluation reports and technical memos to support deployment.
    • Prompt engineering
    • Fine-tuning
    • Evaluation
  4. 2 yrs 10 mos
    • AI & Product
    • Growth & Partnerships
    Mar 2022 — Dec 2024

    Founder, CEO & ML Engineer

    GT Augment · Abuja, Nigeria

    Founded and ran an applied-AI company building education, marketing and augmented-reality products.

    • $100kfunding secured
    • 1,000+users
    • 3countries
    • Developed 3 proprietary AI models for education, marketing and AR-based engagement.
    • Secured $100,000 in funding and scaled across 3 countries to 1,000+ users.
    • Oversaw technical content, investor decks and platform documentation.
    • Founder
    • Applied AI
    • AR
    • Fundraising
  5. 4 mos
    • AI & Product
    Jun 2024 — Sep 2024

    AI Research Intern

    Dugree · New York, USA

    Research on large language model efficiency, contributing to publications and internal symposiums.

    • +45%model efficiency
    • 5+publications
    • Optimised LLMs for a 45% increase in model efficiency.
    • Contributed to 5+ research publications and presented findings at internal symposiums.
    • Authored research briefs and experimental documentation for cross-functional review.
    • LLM optimisation
    • Research
    • Publications
  6. 5 mos
    • AI & Product
    • Teaching
    Mar 2024 — Jul 2024

    AI/ML Research Assistant

    Moscow Institute of Physics and Technology (MIPT) · Moscow, Russia

    Built AI-driven educational models and co-authored academic work on AI in education.

    • +60%student performance
    • Built AI-driven educational models that improved student performance by 60%.
    • Co-authored academic publications and presented at AI education conferences.
    • Wrote technical specifications and user-behaviour analysis summaries.
    • Applied ML
    • EdTech
    • Academic writing

Impact

What the models did.Each figure from the role behind it.

Dugree

+45%

model efficiency

LLM optimisation research in New York.

GT Augment

$100k

funding secured

Raised for the applied-AI company he founded and ran.

MIPT

+60%

student performance

From AI-driven educational models at MIPT.

Dugree

5+

publications

Research contributions, presented at internal symposiums.

Case studies

Selected work.Problem, approach, outcome.

Three projects where the constraint was real and the result is measurable.

  1. 2024Berom Community in Diaspora

    Berom AI

    Researcher & Engineer

    • 800k+speakers served
    • +30%online presence

    Problem

    Berom is spoken by roughly 800,000 people in central Nigeria and is almost absent from the internet. Low-resource languages like it are invisible to the models that increasingly mediate access to information — which means the language quietly stops being usable in the places people now live.

    Approach

    Built an NLP model on transformer architectures trained over linguistically enriched Berom datasets, working with the community in diaspora to assemble material that did not previously exist in machine-readable form.

    Outcome

    Improved accuracy and scalability for low-resource NLP, raised Berom's online presence by 30%, and made the language usable in educational and communicative software for its 800,000+ speakers. The work fed into wider research frameworks on indigenous languages.

    • Transformers
    • Hugging Face
    • PyTorch
    • Low-resource NLP

  2. 2024Ftizisbiomed

    Fluorogram Analysis Model

    ML Engineer

    • 10,000+images processed
    • +20%diagnostic precision

    Problem

    Thoracic abnormalities are missed in routine fluorographic screening, and radiologist time is the bottleneck that makes review at population scale impractical.

    Approach

    Designed a machine-learning pipeline processing 10,000+ fluorographic images using computer vision techniques, tuned for the failure modes that matter clinically rather than for headline accuracy.

    Outcome

    A 20% increase in diagnostic precision, presented at Ftizisbiomed's symposium on AI in healthcare.

    • Computer vision
    • Python
    • TensorFlow
    • Medical imaging

  3. 2024Latoken

    Latoken AI Assistant

    Engineer — hackathon winner

    • −50%response time
    • 1,000+weekly interactions

    Problem

    High-volume customer service where response latency, not answer quality, was the thing losing people.

    Approach

    Engineered a conversational assistant on LLMs with context-aware dialogue management, built for sustained throughput rather than demo conditions.

    Outcome

    Cut response time by 50% while handling 1,000+ interactions a week. Won the Latoken AI Hackathon. The findings contributed to frameworks for scalable AI-assisted customer engagement.

    • LLMs
    • RAG
    • LangChain
    • Dialogue management

Tools

Named, not rated.What the work above shipped with.

  • Machine learning

    AI & Product

    • PyTorch
    • TensorFlow
    • Scikit-learn
    • Hugging Face
    • Transformers
    • Weights & Biases

  • Language models

    AI & Product

    • Fine-tuning
    • Prompt engineering
    • RAG
    • Embeddings
    • LangChain
    • LLaMA 3

  • Engineering

    AI & Product

    • Python
    • SQL
    • Docker
    • Git
    • Airflow

  • Languages

    Every track

    • English — Native
    • Russian — C1 Advanced
GT StudioLooking for film, design or web?The studio is through here — 300+ visual projects, films you can play.Visit the studio