Projects Portfolio

Centific operates as one of Microsoft's largest offshore delivery partners, managing 40+ concurrent AI and ML programs. When I joined as AVP Engineering, delivery was fragmented -- no standardised AI governance checkpoints, inconsistent quality frameworks, and manual-heavy program management creating unnecessary risk across the portfolio.

Design and operationalise a new delivery operating model for 40+ AI programs simultaneously -- embedding responsible AI checkpoints, automated quality gates, and cross-team escalation protocols -- without disrupting active delivery commitments to Microsoft clients.

Architected a new AI delivery governance framework with structured checkpoints (bias reviews, privacy audits, model accountability gates) mapped to each program milestone. Introduced AI-driven automation pipelines replacing manual coordination workflows, reducing overhead by 40%. Built and scaled the delivery team to 450+ professionals with structured enablement covering responsible AI principles and quality standards. Established SLA compliance dashboards and incident resolution tracking that reduced resolution time by 15% YoY.

98% on-time delivery rate and zero major escalations across 40+ programs. 99.5% SLA compliance. 30% reduction in defects. $3.5M in new revenue through strategic account expansion. Reduced manual effort by 40% through AI-driven automation.

Microsoft Purview is the enterprise compliance and data security platform used by Fortune 500 companies globally. Security teams were drowning in alert volume, policy complexity, and manual eDiscovery review -- the product had powerful capabilities but required deep expert knowledge to operate effectively. LLMs offered a path to making the platform accessible to every analyst, not just the top 10%.

Define, build, and ship a suite of LLM-powered Copilot experiences across three distinct security workflows -- eDiscovery summarisation, DLP policy insights, and natural language to keyword query -- while managing the responsible AI, privacy, and compliance constraints of processing sensitive enterprise data at scale.

Owned the full product strategy, roadmap, and delivery for the Purview Copilot suite. Ran customer discovery with security analysts, legal teams, and compliance officers to map the highest-value LLM intervention points in their workflows. Defined the RAG architecture, prompt design standards, and evaluation metrics for each Copilot experience. Partnered with Microsoft Legal and Privacy to establish the responsible AI boundaries for LLM access to sensitive customer data. Resolved LLM performance bottlenecks with engineering -- reducing response time by 25% across pilot tenants.

16K+ monthly Copilot invocations at GA. 30% faster security triage for analyst teams. $3.2M/year in measurable productivity gains. 50% reduction in manual policy authoring effort. Shipped eDiscovery NLS Copilot, DLP Policy Insights Copilot, and NL-to-KQL -- all in production.

Enterprise security teams configuring Data Loss Prevention policies had no visibility into whether their policies were actually working. Sensitive data could flow through gaps in policy coverage for months without detection. The problem wasn't policy enforcement -- it was that organisations didn't know what they didn't know.

Build a product that could proactively surface policy blind spots and misconfigured DLP rules across enterprise tenants -- without requiring analysts to know where to look. The technical challenge was building anomaly detection across highly varied customer environments with different data types, industries, and threat profiles.

Defined the product concept and wrote the PRD for the DLP Analytics Engine -- an ML-based diagnostic that analysed data flows across enterprise tenants and flagged deviations from expected policy coverage patterns. Led the data science collaboration to select and tune the anomaly detection model. Designed the UX for surfacing insights in the analyst workflow -- specifically ensuring recommendations were actionable, not just informational. Managed the $5M portfolio with 50+ global vendors contributing telemetry and signal data.

20% month-over-month growth in Monthly Protected Users (MPU) -- the platform's north star metric. Surfaced systematic policy blind spots that had been invisible to enterprise security teams. Established the data-driven foundation for subsequent DSPM Agent work.

Enterprise security operations centres were facing a structural analyst shortage. Alert volumes were growing faster than headcount. The existing tooling required a human analyst to review, contextualise, and resolve every security alert -- a workflow that didn't scale. DSPM, DLP, and IRM each had their own alert queues, their own resolution workflows, and their own expert knowledge requirements.

Design and ship autonomous AI agents that could triage, contextualise, and propose resolutions for security alerts across three distinct product domains (DSPM, DLP, IRM) -- while defining the governance model for when agents act autonomously versus when they escalate to a human analyst.

Led product strategy and roadmap for all three agents simultaneously. Defined the agent architecture: tool calling design, context retrieval strategy, confidence scoring, and human-in-the-loop escalation thresholds. Built the governance framework specifying what actions agents could take autonomously (low-risk recommendations) versus what required human review (policy changes, data access decisions). Designed the feedback loop for continuous agent improvement using analyst accept/reject signals as RLHF training data.

30% reduction in analyst workload across all three agent products. $2M/year in measurable productivity gains. 30% improvement in resolution accuracy versus unassisted analyst triage. Three production agents shipped -- DSPM, DLP Alert Triage, and IRM Alert Triage.

Mastercard's B2B Track platform aimed to digitise and streamline B2B payment flows for corporate buyers and suppliers globally. The core challenge was entity resolution -- matching buyer and supplier identities across fragmented, inconsistent supplier master data at global scale. Incorrect entity matching meant failed payments, delayed reconciliation, and broken supply chain finance flows.

Architect and deliver the Matching Engine and Entity Resolution system for Mastercard's B2B Track platform -- a distributed system that had to reliably identify and match corporate entities across geographies, naming conventions, and data quality levels while meeting the strict uptime and compliance requirements of financial infrastructure.

Led end-to-end architecture and delivery across distributed teams in the US, India, and Europe. Drove requirements analysis and solution design for the Matching Engine -- defining the entity resolution rules, confidence scoring approach, and exception handling for ambiguous matches. Acted as the primary liaison between Mastercard's product teams and the Microsoft delivery organisation, ensuring technical decisions were aligned with business requirements across time zones and cultures.

On-time delivery of Mastercard's B2B Track platform across all three geographies. Matching Engine and Entity Resolution delivered and operational. Established the data governance patterns for a platform now used across Mastercard's global B2B network.

Dubai Electricity & Water Authority (DEWA) serves millions of customers across Dubai who interact with the utility in Arabic and English. Their customer service operation was handling enormous volumes of routine queries -- billing, fault reporting, service requests -- through human agents. The vision was an AI-first customer service layer that could handle the majority of these interactions autonomously at any hour.

Architect and deliver a production multilingual AI chatbot capable of handling 24/7 customer interactions for a critical utility -- with Arabic NLP quality requirements, real-time uptime SLAs, and a self-learning mechanism that continuously improved from unanswered queries without requiring manual retraining cycles.

Led requirements analysis, stakeholder alignment across DEWA and Microsoft teams, and the end-to-end deployment strategy. Designed the self-learning feedback loop: queries the bot couldn't answer were flagged, reviewed, and used to extend the training corpus -- creating a continuous improvement cycle without manual intervention. Drove the multilingual NLP architecture to handle Arabic morphological complexity and dialect variation alongside English. Managed the deployment strategy to ensure zero downtime transitions in a 24/7 utility context.

35% improvement in response accuracy within 6 months of deployment. 24/7 autonomous customer interaction capability established. Enabled DEWA to handle significantly higher query volumes without proportional headcount growth.

Two organisations -- Miami Dolphins (NFL franchise) and GPG Education (South Africa) -- needed real-time reporting and analytics platforms built on Azure. For the Dolphins, the constraint was game-day: analytics had to be live during matches, not available the next morning. For GPG, the challenge was education performance tracking across schools with inconsistent data quality.

Design and deliver a real-time analytics platform integrating Azure SQL, Power BI, Dynamics CRM, Ticketmaster, and Azure DevOps data streams -- with a CI/CD pipeline that could support rapid feature iteration without breaking live reporting during high-stakes game-day events.

Led full architecture and delivery for both implementations. For the Dolphins: designed the streaming data ingestion from Ticketmaster and CRM systems into an Azure SQL analytical layer, with Power BI dashboards refreshing on near-real-time cycles. Implemented CI/CD automation with quality gates to enable fast deployments without production risk. For GPG Education: adapted the platform for education metrics and South African data infrastructure constraints.

90% reduction in deployment cycle time through CI/CD automation. Real-time game-day analytics capability delivered for the Dolphins. Education performance analytics platform operational across GPG schools in South Africa.

Tanla Platforms (now a publicly listed Indian telecom platform company) needed a blockchain-based system to bring trust and transparency to business messaging transactions -- verifying that commercial SMS messages from enterprises to consumers were legitimate, traceable, and tamper-proof. This was a regulatory and trust infrastructure problem as much as a technology one.

Architect a blockchain-based transaction platform capable of handling 5M+ daily SMS transactions with high throughput, low latency, and immutable audit trails -- meeting TRAI regulatory requirements for enterprise messaging transparency in India.

Led the full architecture design for Trubloq -- selecting the distributed ledger architecture, designing the transaction schema and consensus mechanism appropriate for the throughput requirements, and defining the integration model with Tanla's existing messaging infrastructure. Drove the scalability design to handle 5M+ daily transactions reliably with the audit trail guarantees the regulatory context required.

Platform operational handling 5M+ daily transactions. Immutable audit trail established for enterprise messaging compliance. Tanla Trubloq became a foundational part of India's TRAI-compliant business messaging infrastructure.

Enterprise customers running Hyper-V virtualisation faced a critical security gap: VM administrators with privileged access could inspect or tamper with guest VMs containing sensitive workloads. Shielded VMs addressed this by cryptographically protecting VMs from the fabric administrators who managed the infrastructure -- a significant advance for regulated industries.

Define, build, and launch a net-new security feature -- Hyper-V Shielded VM in SCVMM 2016 -- that addressed a genuine enterprise security concern while navigating the complexity of a feature deeply embedded in the virtualisation stack. Then drive adoption in a market where IT administrators are historically resistant to change.

Owned the full 0-to-1 product lifecycle: defined requirements with engineering teams, specified the security model and compliance use cases, and managed the release including critical hotfix releases post-GA. Developed the go-to-market strategy: produced Channel9 demo videos, created marketing and sales enablement materials, and built the data-backed insights that helped field teams articulate the security value proposition to enterprise buyers.

Launched Hyper-V Shielded VM in SCVMM 2016 with 40% boost in enterprise adoption. Feature adopted across regulated enterprise segments including finance and healthcare.

Microsoft Consulting Services ran 5,000+ client engagement estimates per year globally -- each one produced by individual consultants using their own spreadsheets, tribal knowledge, and inconsistent assumptions. The result was massive estimation variance, missed bids, and no institutional learning from past engagements. The problem wasn't effort, it was fragmentation.

Replace the entire estimation culture across Microsoft Consulting Services globally -- across 7 countries and multiple consulting practices -- with a standardised platform that consultants would actually adopt. The hard part wasn't building the tool; it was getting 5,000+ experienced consultants to stop doing it their way.

Led the full program as Group Product Manager: defined the product from first principles (not digitising the spreadsheet, but redesigning the estimation model), ran stakeholder alignment across domain architects and practice leads across Europe, and orchestrated the change management campaign that drove adoption. Built the coalition of early adopters in each country who became internal champions. Designed the PROSCI-informed adoption strategy -- awareness, training, reinforcement -- that moved the adoption needle from pilot to 75% global rollout.

$4M+ in estimated annual productivity savings. 75% global adoption across Microsoft Consulting Services. 5,000+ estimations per year now running through the platform. Became a core tool for all consultants in sales and pre-sales estimation.

Microsoft Global Delivery needed to develop the next generation of senior technical leaders across its European consulting practices -- architects and program managers who could lead complex enterprise engagements independently without escalation. The existing training programs were generic and didn't address the specific leadership challenges of technical delivery in a client-facing, cross-cultural consulting environment.

Design and deliver a bespoke technical leadership program from scratch -- across seven countries, in partnership with CTOs and senior leaders who were sceptical of 'corporate training' -- and sustain it at scale for three consecutive years while continuously improving based on participant feedback.

Acted as Program Director: designed the full curriculum, secured executive sponsorship from regional CTOs, and built the delivery partnership with senior leaders who became faculty. Ran cross-regional change management to align each country's leadership team with the program objectives. Applied data-driven iteration -- using NSAT scores and cohort feedback to refine content and delivery format each year.

NSAT > 180 for three consecutive years. Program scaled across 7 countries. Strengthened Microsoft's senior leadership pipeline across European consulting. Recognised as a flagship talent development initiative within Microsoft Global Delivery.

Fortune 500 clients running legacy SharePoint and on-premise infrastructure needed to modernise to SharePoint 2013 and Azure-based Hyper-V environments -- while maintaining 24/7 intranet and extranet continuity for thousands of enterprise users. Any downtime during migration had direct business impact.

Deliver complex infrastructure modernisation programs for Fortune 500 clients across the US and Denmark -- managing cross-regional delivery teams, coordinating with on-site client stakeholders, and executing zero-downtime migrations including a complex DB2-to-SQL-Server data migration for Praxair with full data integrity requirements.

Led full project lifecycle across multiple concurrent enterprise clients. Managed cross-regional teams spanning onsite client locations and offshore delivery. Designed the migration sequencing to achieve zero downtime -- running parallel environments through cutover windows planned around client operational rhythms. Executed the DB2-to-SQL-Server migration for Praxair with a validation framework that verified data integrity at every stage before decommissioning source systems.

35% improvement in platform reliability post-migration. 30% reduction in infrastructure operating costs. Zero downtime across all client migrations. Full data integrity maintained through the Praxair DB2-to-SQL migration.

AMC Theatre's Point of Sale systems needed to be integrated with their Enterprise Data Warehouse -- a mission-critical integration that underpinned sales reporting, inventory management, and financial reconciliation across hundreds of theatre locations. The existing BizTalk 2006 infrastructure then required a major version migration to BizTalk 2010 without disrupting the live integrations.

First: architect and deliver the EAI integration between AMC's PoS systems and EDW using BizTalk Server -- a complex, mission-critical system with real-time transaction flows. Second: execute the first-ever Global Delivery migration from BizTalk 2006 to BizTalk 2010, integrating EDI, Microsoft Dynamics AX, Oracle, and external vendor systems -- all with zero downtime tolerance.

Acted as on-site coordinator in the USA: led requirements gathering with AMC's business and IT teams, managed the EAI architecture and deployment, and conducted user training for adoption continuity. For the BizTalk migration: designed the migration sequence, built the integration test framework to verify each system connection before cutover, and managed the cross-system dependencies between EDI, Dynamics AX, Oracle, and third-party vendors.

EAI system operational -- PoS-to-EDW data flowing reliably across all AMC locations. BizTalk migration completed with zero downtime -- the first successful Global Delivery migration of its kind. All five enterprise systems (EDI, Dynamics AX, Oracle, PoS, EDW) integrated and operational post-migration.

Baxter Healthcare needed a clinical software application -- RenalSoft -- for patients using their home dialysis equipment. This was Class II medical device software under FDA jurisdiction, used directly by patients (not clinicians) in their homes. The stakes were unambiguous: software defects could directly harm patients.

Develop v1.0 of a mission-critical FDA-regulated clinical software application for home dialysis -- meeting FDA 510(k) design control requirements, navigating the validation and verification documentation standards, and designing for a user who is a home patient, not a clinical professional.

Led the full development of RenalSoft v1.0: requirements definition, software design, implementation, and verification against FDA design control standards. Designed the software with patient safety as the primary design constraint -- every interaction model, error state, and data entry flow reviewed against the risk that a patient in a home setting would encounter it without clinical supervision.

RenalSoft v1.0 delivered and FDA-compliant. Clinical software operational for Baxter's home dialysis patient population. Established the regulatory documentation and design control framework that would support future versions.