With a background spanning software development, data engineering, data science, analytics, and project management & leadership roles
10+ years of professional experience across software engineering, data analytics, data science, product management, and team leadership — spanning industries from healthcare to high-scale consumer platforms. I bring genuine technical depth together with the product strategy and people skills to take ideas from 0→1 all the way to enterprise scale. Now focused on AI product and engineering leadership roles where both sides matter.
Click a node or label to explore
From data science and deep learning to LLMs, data engineering, and HIPAA-compliant cloud architecture — I don't just manage AI products, I understand every layer that makes them work. That foundation in both research and engineering is what allows me to deliver AI products that are secure, scalable, and truly built to last.
First-ever Director of PM at Theoria Medical, where I built the entire product management function from the ground up. Owned product strategy and roadmaps tightly aligned to business goals, translating C-suite priorities into clear execution across both B2B and B2C products. Led the integration of multiple third-party platforms and vendor systems into a single, seamless product experience, and brought hands-on technical depth to large-scale cloud migration across Azure, GCP, and AWS. Working as the bridge between leadership and the engineering, data, and clinical teams, I drove a 300% increase in hiring output through automation and cut product defects by 50% with structured QA governance — leading cross-functional agile teams of 10+ across discovery, prioritisation, and delivery.
Led and mentored a team of data engineers and Python developers, guiding architecture decisions and engineering best practices across the data platform. Drove cloud infrastructure optimisation on GCP BigQuery and Microsoft Azure for stronger reliability and pipeline performance, and built scalable API-driven, Apache Airflow–orchestrated platforms integrating third-party healthcare systems. Established HIPAA-compliant data governance frameworks that improved data trust, regulatory compliance, and executive reporting accuracy.
Designed and deployed scalable ETL pipelines and cloud data infrastructure on GCP and Azure to support healthcare operational reporting and analytics. Integrated third-party data systems via REST APIs, improving data reliability and reducing manual processing overhead.
Built ML models (XGBoost, Random Forest, GLM) to optimise pricing and trading strategies on scalable Azure and MongoDB platforms processing large datasets at one of the world's sharpest sportsbooks. Partnered with quant teams to improve risk management systems and delivered data-driven product insights to enhance trading tool UX.
Led analytics product modernisation for Sales and Marketing, transitioning manual Excel workflows to automated cloud-based ML pipelines. Delivered predictive sales forecasting models supporting financial planning and quota strategy. Collaborated with global stakeholders to align analytics with regional business goals.
Managed full SDLC for embedded systems software. Translated enterprise client requirements into technical specifications and contributed to patent-generating innovations in hardware-software integration.
A mix of personal AI product concepts — shown with how I'd take them to market as a PM — and hands-on research projects across healthcare, iGaming, and data science.
A personal product concept: one platform that gives each person exactly the data they need, on web or mobile. Providers see and manage their own schedule, pay outlook, and compliance; clinic managers see across the whole team; engineers get a technical view. Same data, role-based access, built to live inside the EHR or as a standalone app. The goal: more accountable, more compliant providers — and less manual coordination. Tap below for the full product brief.
A personal product concept for a data-driven odds-optimisation tool that helps a trading desk tighten pricing, reduce margin leakage, and react faster to market moves. As the PM, I'd map the trading workflow and its real pain points, prioritise a single high-value sport for an MVP model, and define guardrails and success metrics together with traders and risk before scaling across markets.
A personal product concept for a responsible-gambling tool that flags at-risk behaviour from deposit velocity, session length, and bet-sizing patterns, then triggers timely, regulation-aligned interventions. As the PM, I'd work with compliance and support teams to define risk thresholds and intervention playbooks, ship a monitored MVP in one market, and tune for fairness and false-positive rate before a wider rollout (EU DGOJ / UK GC).
Novel CNN-BiLSTM hybrid models detecting depression from ECG physiological signals. Applied STFT spectrogram transformation, Neurokit2 pre-processing, and VGG16/ResNet50 transfer learning. Used CNN-LSTM, CNN-BiLSTM, VGG16-LSTM, VGG16-BiLSTM, and ResNet50-BiLSTM architectures.
Automated malaria detection from microscopic cell images combining ML and DL. Used Support Vector Machines (SVM) alongside Convolutional Neural Networks (CNN) trained on annotated infected vs uninfected cell image datasets. Libraries: NumPy, Pandas, TensorFlow, Keras, PyTorch.
Multi-model classification comparison using XGBoost, Decision Tree, Random Forest, SVM, and GLM in R. Applied Boruta and Recursive Feature Elimination (RFE) for feature selection with 10-fold cross-validation. Used EDA, feature engineering, and metrics including ROC_AUC and F1-Score.
Rare full-stack capability across product leadership, data engineering, ML/AI, and cloud infrastructure.
Download or view my full CV: all experience, skills, and education in one document.
Seeking senior PM, Head of Product, and AI Product Leadership roles in healthcare tech, iGaming, and data-intensive industries. I bring both the strategic vision and the technical credibility to lead and deliver.
Everything healthcare in one place — my roles at Theoria Medical, a healthcare product concept, and hands-on ML projects on clinical data.
Built the product management function from the ground up — business-aligned roadmaps, B2B & B2C products, third-party integrations, and hands-on cloud migration across Azure, GCP, and AWS. Drove 300% hiring output and cut defects by 50%.
Led and mentored data engineers and Python developers; built scalable, Airflow-orchestrated platforms integrating third-party healthcare systems, with HIPAA-compliant data governance.
Built and maintained data pipelines and integrations across clinical, scheduling, and billing systems, supporting analytics and reporting in a HIPAA-regulated environment.
One platform that gives each person exactly the data they need — providers manage their own schedule, pay outlook, and compliance; managers see the team; engineers get a technical view. Role-based access, built to live inside the EHR or as a web & mobile app.
Novel CNN-BiLSTM hybrid models detecting depression from ECG physiological signals. Applied STFT spectrogram transformation, Neurokit2 pre-processing, and VGG16/ResNet50 transfer learning.
View on GitHub →Automated malaria detection from microscopic cell images combining ML and DL — Support Vector Machines (SVM) alongside Convolutional Neural Networks (CNN) trained on infected vs uninfected cell images.
View on GitHub →One trusted source of data, shown through the right lens for each person — on web or mobile. Everyone sees exactly what they need to do their job, and nothing they shouldn't, whether it lives inside the EHR or as a standalone app.
Clinics already collect the data they need — it's just scattered across the EHR, scheduling, billing, and other tools, and no one sees a complete picture relevant to their own job. ClinIQ brings that same data together once, then shows each person only the slice that's useful and allowed for them. A core goal: let providers manage their own schedule directly — see what's compliant, fix what isn't, and stay on top of their own visits — instead of relying on someone else to do it for them. The payoff is more accountable, more compliant providers and less manual coordination behind the scenes.
From scattered data to confident, compliant action — at a glance:
Everyone works from the same trustworthy data, but each role only sees what's relevant and permitted to them, set by access levels. A provider sees their own world; managers see across the whole team; engineers see the technical layer. Nobody sees data they shouldn't.
AI is what turns the unified data from a static report into something that looks ahead. It always assists — people stay in control of the final decision.
Under the hood: forecasting uses regression / gradient-boosted models, the risk flags use classification, anomaly detection uses unsupervised methods, and plain-English questions use an LLM — all hosted on the cloud's managed ML service, with monitoring to catch model drift over time.
Cloud-flexible by design — it runs on whichever platform the company already uses. The pieces map cleanly to either Google Cloud (GCP) or Microsoft Azure:
Healthcare data is sensitive and the rules differ by region — for example HIPAA in the US and GDPR in Europe, among others. Security is built in from the start:
ClinIQ is designed to sit inside the existing EHR and/or as a standalone web and mobile app. The data behind both is identical — only the way each person reaches it differs — so people can use it without leaving the tools they already work in, and providers can manage their schedule right from their phone.
In: one unified source of data, the full provider view (schedule with compliance flags, 3-month pay forecast, missed and non-compliant visits), web and mobile friendly, and role-based access for providers, managers, and the technical team.
Out (for now): advanced predictive analytics, every outside integration, plain-English search, and multi-region rollout — added once the core earns trust.
Assumption: the underlying scheduling, EHR logic, and integrations already work reliably — ClinIQ sits on top of them and reads from them.
A lean cross-functional squad of about 6–7 people, scaling up only once the MVP proves its value:
Part-time & advisory: a compliance & privacy advisor (HIPAA / GDPR), a clinical subject-matter expert (a real provider representing day-to-day needs), and DevOps support for secure deployment.
Stakeholders kept in the loop: providers, clinic managers, compliance / legal, and engineering leadership — consulted in discovery and at each release.
A PM should give a credible target even before everything is known. My indicative estimate for the MVP (provider view + role-based access) is roughly 3–4 months:
This is a planning estimate, not a promise — it firms up after discovery and depends on data access and the chosen cloud.
Phase 1 — Discovery: sit with providers, managers, and engineers to confirm exactly what each role needs to see. Phase 2 — MVP: ship the provider view with role-based access in one clinic and region. Phase 3 — Scale: add the manager and developer views, more integrations, and additional regions once the value is proven.
Personal product concept — illustrative of how I'd approach this as a PM, not a shipped product.
A data-driven tool that helps a trading team price smarter and react faster — tighter odds, less margin lost, and clearer decisions, on web or mobile.
A sportsbook lives or dies on its pricing. A lot of that still leans on manual judgment and slow tools, so prices can lag the market and margin quietly leaks away. This concept brings the relevant data together and lets AI suggest sharper prices, while traders stay firmly in control — they review, adjust, and approve. The goal: faster, more confident pricing with less guesswork.
From raw market data to a confident price — at a glance:
Everyone works from the same trusted data, but each role sees only what's relevant and permitted: traders see their markets, managers see across the desk, engineers see the technical layer. Nobody sees data they shouldn't.
AI assists — it proposes, people decide.
Under the hood: gradient-boosted models (e.g. XGBoost) and time-series methods for pricing, anomaly detection for risk flags — all on the cloud's managed ML service, with monitoring to catch drift.
Cloud-flexible — runs on whichever platform the company uses, mapping cleanly to Google Cloud (GCP) or Microsoft Azure:
Betting is heavily regulated and the rules vary by market — for example the UKGC in the UK and the MGA in Malta, among others. Safeguards are built in from the start:
A trader cockpit on web and mobile, with the option to plug into the existing trading platform. Same data, surfaced where the team already works.
In: one sport / market, the data pipeline, a first pricing model with trader review-and-approve, and role-based access.
Out (for now): fully automated pricing, every market, and plain-English search — added once the core earns trust.
Assumption: the live data feeds and trading platform already work reliably — the engine sits on top of them.
A lean cross-functional squad of about 6–7 people, scaling up only once the MVP proves its value:
Part-time & advisory: a compliance advisor (licensing / data rules), a trader subject-matter expert, and DevOps for secure deployment.
Stakeholders kept in the loop: traders, trading / risk managers, compliance, and engineering leadership.
A planning estimate for the MVP (one market, review-and-approve) — roughly 3–4 months:
A planning estimate, not a promise — it firms up after discovery and depends on data access and the chosen cloud.
Phase 1 — Discovery: map the trading workflow and its real pain points. Phase 2 — MVP: ship one market with review-and-approve. Phase 3 — Scale: add markets and more automation once trust is proven.
Personal product concept — illustrative of how I'd approach this as a PM, not a shipped product.
A safer-gambling tool that spots signs of harm early and helps the team step in with the right support at the right time — fairly, and within the rules.
Most signs of harm are already in the data — sudden jumps in deposits, very long sessions, chasing losses — but they're easy to miss until it's too late. This concept watches for those patterns and surfaces them early, so a safer-gambling team can reach out with support before harm deepens. People always make the final call; AI just helps them see sooner.
From quiet warning signs to timely, caring support — at a glance:
Everyone works from the same trusted data, but each role sees only what's relevant and permitted — and player privacy is respected at every level. Nobody sees data they shouldn't.
AI assists — it surfaces concerns, people decide how to help.
Under the hood: classification and anomaly detection on behavioural signals, always with a human in the loop and ongoing fairness and accuracy monitoring.
Cloud-flexible — maps cleanly to Google Cloud (GCP) or Microsoft Azure:
This is highly sensitive data, governed by both gambling regulators — e.g. the UKGC and MGA — and data-protection law such as GDPR. Safeguards are built in from the start:
A case-management dashboard on web and mobile, able to plug into the existing platform. Same data, surfaced where the team already works.
In: a behavioural-data pipeline, a first risk model with human review, a simple case / intervention workflow, role-based access, and one market.
Out (for now): fully automated interventions, every signal source, and multi-region rollout — added once the core earns trust.
Assumption: the underlying player-data feeds already work reliably — the monitor sits on top of them.
A lean cross-functional squad of about 6–7 people, scaling up only once the MVP proves its value:
Part-time & advisory: a safer-gambling / compliance expert, a support-team subject-matter expert, and DevOps for secure deployment.
Stakeholders kept in the loop: the safer-gambling team, support, compliance / legal, and engineering leadership.
A planning estimate for the MVP (one market, human-in-the-loop) — roughly 3–4 months:
A planning estimate, not a promise — it firms up after discovery and depends on data access and the chosen cloud.
Phase 1 — Discovery: define risk signals and intervention playbooks with the safer-gambling and support teams. Phase 2 — MVP: ship one market, human-in-the-loop. Phase 3 — Scale: add more signals and markets, and automate the routine steps, once trust is proven.
Personal product concept — illustrative of how I'd approach this as a PM, not a shipped product.