Download PDFDownload WordJSON-LD dataSFIA mapping

Dr Alexander Mikhalev

Online CV · JSON-LD data

Managing Partner and CTO | Applied AI and Autonomous Delivery

Swindon, United Kingdom

LinkedIn · GitHub · metacortex.engineer

Profile

Managing Partner, CTO and Head of AI combining technology direction with hands-on engineering in Rust, Python and TypeScript. Builds AI capability and the software platforms that put it to work: industrial monitoring and anomaly detection, knowledge search, client applications and autonomous engineering infrastructure. Experience spans startup team building, enterprise architecture at Nationwide and applied research at Cranfield. Leads delivery while remaining close to the code, data and engineering decisions.

Current leadership

Zestic AI

Managing Partner | CTO and Head of AI | July 2023–present

Platforms built and operated

Terraphim AI

Architecture and hands-on development | Independent open-source platform

terraphim.ai

Autonomous delivery factory

AI Dark Factory and AgentForge

Other concurrent appointments

Solutions Architect, InfinyOn | Contract | December 2022–December 2023

Director, Applied Knowledge Systems | Self-employed | December 2022–present

Industrial AI leadership

Klarian

Chief Technology Officer | November 2023–June 2025

AI/ML Architect | December 2022–December 2023

Earlier leadership

Nationwide Building Society

Architecture and prototyping leadership | December 2014–December 2022

AI/ML Architect, January 2019–December 2022; Head of Prototyping Engineering, January 2018–January 2019; Solutions Architect, Accelerated Change Efficiency and DevOps, January 2017–January 2018; Tech Lead, December 2014–January 2017.

M2M Ecosphere

Director of R&D | January 2019–January 2021 | Concurrent appointment

Shopitize

Head of Architecture and Development | November 2011–December 2014

Research and specialist delivery

Cranfield University

Research Fellow | May 2007–January 2012

Selected applied AI engagements

IT/IS Executive | Microsharp Corporation | February 2003–May 2007

Designed and tested company IT and communications infrastructure with automation and redundancy.

Technical foundation

Platform engineering: Rust, Python, TypeScript, APIs, distributed systems, PostgreSQL, Redis, containers, CI/CD and Cloudflare Workers. Historical delivery also includes Java/Spark, TensorFlow, PyTorch and MATLAB.

Applied AI and systems: machine learning, NLP, knowledge graphs, graph embeddings, search, image processing, sensor fusion, privacy-preserving approaches and synthetic data. Systems engineering, technical architecture, prototyping and delivery governance connect the research and implementation work.

Patents

Secure distributed data storage | UK application GB1815423.7

Granted patent GB2574076B. Led the prototyping team and supported patenting; sponsored and mentored the named inventor.

Loyalty-program technology | Patent application WO2013057491

Named co-inventor of “Method and system for providing a loyalty program” for Shopitize.

Awards

Selected talks

Competition projects and publication

MedGemma competition, February 2026: built a solo knowledge-grounded research prototype in Rust, including entity extraction, a knowledge graph, multi-agent orchestration, MedGemma integration and an evaluation harness; evaluated across 18 scripted cases with CPU/GPU inference.

Industrial publication: Co-author, Monitoring and Anomaly Detection Approaches with AI and Data Analytics for Pipelines, Pipeline Technology Journal, 2023.

Education

PhD | Cranfield University

Thesis: Image processing and agent-based framework for the geolocation of emitters (2010).

MSc | Bauman Moscow State Technical University

Computer Science and Computer Engineering.

TOGAF 9 Certified | The Open Group

Career themes

The following connections are thematic interpretations of the experience above. They do not assert additional inventions or SFIA ratings.

Digital twins and industrial AI

At Cranfield, researched in-vehicle sensor networks within a Boeing- and Rolls-Royce-backed Integrated Vehicle Health Management (IVHM) programme. This work connects to the sensor-data and modelling foundations of digital twins; later Klarian work applied physical models and machine learning to industrial monitoring.

Graph-based knowledge and grounded AI

The Pattern combined document exploration and graph-based search. Terraphim develops role-specific knowledge graphs and interpretable graph embeddings; the MedGemma research prototype brings knowledge graphs into a grounded AI workflow.

Privacy and enterprise AI capability

At Nationwide, led the Synthetic Data Platform from inception to enterprise capability and developed privacy-preserving machine-learning approaches. Terraphim continues the concern for data control through local-first search and context engineering.

From prototypes to autonomous delivery

Built prototyping capability and architecture-as-code demonstrations at Nationwide, then developed reusable delivery practices and an autonomous delivery factory connecting agent execution with separate review, verification and validation gates.

SFIA skills and behaviour signals

An evidence-informed mapping to SFIA 9, not certification or a formal responsibility assessment. Levels are provisional; upper candidates require further evidence. Behaviour signals are ungraded. Awards, talks and job titles do not independently establish a responsibility level.

Each capability below retains its context, rationale and evidence gaps. Data vocabulary explains the mapping fields.

Profile · profile-01

Managing Partner, CTO and Head of AI combining technology direction with hands-on engineering in Rust, Python and TypeScript. Builds AI capability and the software platforms that put it to work: industrial monitoring and anomaly detection, knowledge search, client applications and autonomous engineering infrastructure. Experience spans startup team building, enterprise architecture at Nationwide and applied research at Cranfield. Leads delivery while remaining close to the code, data and engineering decisions.

Behaviour signals

Mapping rationale: Summary supported by the detailed platform, banking and industrial claims; it introduces no additional level evidence.

Evidence category: mixed_summary; mapping confidence: medium.

Evidence gaps: Assess the detailed rows; do not count this summary again.

Zestic AI · zestic-ai-02
  • Set technology direction and delivery approach across the client portfolio, combining hands-on platform engineering with reusable AI infrastructure and staged engineering reviews.

Behaviour signals

  • Leadership: Sets delivery approach across projects
  • Planning: Stages reviews and reusable infrastructure

Mapping rationale: Portfolio delivery direction and staged engineering reviews fit development leadership and lifecycle practice ownership.

Evidence category: first_party_operating_records; mapping confidence: medium.

Evidence gaps: For SLEN6: confirm organisational commitment, cross-lifecycle policies, risk ownership and evaluation of the working environment; DLMG6 additionally needs technical/financial/quality targets and resource authority.

Zestic AI · zestic-ai-03
  • Shape a services-and-platform strategy that turns client delivery into reusable capabilities; connect architecture choices, pilot scope and delivery priorities to the wider business model.
SFIA skillProvisional levelUpper candidate
Enterprise and business architecture STPL5
Strategic planning ITSP5

Behaviour signals

  • Decision-making: Connects pilot scope to business priorities
  • Planning: Balances services and reusable platform priorities

Mapping rationale: Business-model canvas connects services/platform choices to capability and delivery priorities; ITSP is a candidate only insofar as this informs the whole business strategy.

Evidence category: first_party_operating_records; mapping confidence: medium.

Evidence gaps: Show approved strategy, authority, investment decisions and outcome reviews before claiming enterprise strategy ownership or level6.

Zestic AI · zestic-ai-04
  • Confidential digital-learning provider: lead architecture and engineering across assessment services, web applications and data models; implement assessment evidence handling, learner-journey testing, analytics instrumentation and CI/CD.

Behaviour signals

Mapping rationale: Architecture ownership and implementation across a multi-component client platform support significant solution/software responsibility; CI/CD is narrower lifecycle implementation.

Evidence category: first_party_implementation_records; mapping confidence: medium.

Evidence gaps: Add acceptance evidence, delivery accountability and specific review decisions; records include integration work, not blanket production completion.

Zestic AI · zestic-ai-05
  • Confidential impact-finance lender: engineer borrower workflows, financial assessment, document generation and CRM integration, bringing operational lending processes into a connected software platform.
SFIA skillProvisional levelUpper candidate
Programming/software development PROG4
Software design SWDN4

Behaviour signals

Mapping rationale: Connected borrower workflows, assessment, generated documents and CRM demonstrate complex application/integration construction.

Evidence category: first_party_implementation_records; mapping confidence: medium.

Evidence gaps: Confirm end-to-end technical accountability before level5; no AI underwriting or regulatory certification inferred.

Zestic AI · zestic-ai-06
  • Build TruthForge, a Rust-based narrative-analysis application, with supporting deployment and website delivery.
SFIA skillProvisional levelUpper candidate
Programming/software development PROG4

Behaviour signals

Mapping rationale: Rust narrative application and delivery contribution evidence software construction, not ownership of every lifecycle stage.

Evidence category: first_party_implementation_records; mapping confidence: medium.

Evidence gaps: Provide reviewed contributions, lifecycle ownership and release acceptance for level5.

Terraphim AI · terraphim-ai-02
  • Lead development of a Rust/WebAssembly knowledge-search platform combining role-specific knowledge graphs, deterministic text matching, command-line tools and Model Context Protocol integration.
SFIA skillProvisional levelUpper candidate
Programming/software development PROG5
Solution architecture ARCH5

Behaviour signals

Mapping rationale: Leadership plus implemented Rust/WASM, CLI and MCP components supports technical platform responsibility and integration architecture.

Evidence category: first_party_implementation_records; mapping confidence: medium.

Evidence gaps: Document collaborator responsibilities, roadmap trade-offs and operating outcomes; public source does not prove organisation-wide policy authority.

Terraphim AI · terraphim-ai-03
  • Build local-first, privacy-preserving search and context-engineering capabilities that connect domain knowledge, session history and reusable learning to AI tools.
SFIA skillProvisional levelUpper candidate
Software design SWDN5
Knowledge management KNOW4

Behaviour signals

Mapping rationale: Context search design supports software design; organising domain/session knowledge supports a knowledge initiative, not proof of an organisation-wide knowledge programme.

Evidence category: first_party_implementation_records; mapping confidence: medium.

Evidence gaps: Show knowledge adoption/impact and policies for KNOW5; privacy architecture is not a full ethics or security assurance programme.

Terraphim AI · terraphim-ai-04
  • Develop Terraphim’s distinctive graph-embedding approach in Rust, using curated concepts, synonym mappings and graph co-occurrence to rank documents and expose the relationships behind search results.
SFIA skillProvisional levelUpper candidate
Programming/software development PROG5
Software design SWDN5

Behaviour signals

Mapping rationale: Inspected Rust graph code ranks documents using concepts, synonyms and co-occurrence; this is concrete software/design evidence, not by itself learned-model MLNG evidence.

Evidence category: first_party_inspected_source; mapping confidence: medium.

Evidence gaps: Add design decisions and comparative evaluation; no uniqueness, patent or performance superiority level inference.

Autonomous delivery factory · autonomous-delivery-factory-02
  • Built and operate the AI Dark Factory, connecting issue prioritisation and agent execution with separate review, verification and validation gates across development projects.

Behaviour signals

Mapping rationale: Implemented operating pipeline joins issue workflow with independent stages of engineering assurance.

Evidence category: first_party_deployment_records; mapping confidence: medium.

Evidence gaps: For SLEN6 confirm organisational commitment and integration of lifecycle policy, risk management and evaluation; deployed cross-project working environment is a candidate6 signal. DLMG stays5 pending resource/target evidence; no productivity gains inferred.

Autonomous delivery factory · autonomous-delivery-factory-03
  • Extended AgentForge, a Gitea-based development platform, with scoped agent identity, signed audit events, a unified event stream and branch-based collaboration.
SFIA skillProvisional levelUpper candidate
Programming/software development PROG5
Software design SWDN5

Behaviour signals

Mapping rationale: Implemented AgentForge identity, audit and event extensions demonstrate complex platform changes and design controls.

Evidence category: first_party_implementation_records; mapping confidence: medium.

Evidence gaps: Security mechanisms do not establish SCTY governance or independent security assurance; confirm contribution scope and lifecycle accountability.

Autonomous delivery factory · autonomous-delivery-factory-04
  • Connected the development platform to a kanban workflow through a Rust bridge, making task coordination and review part of the engineering system.

Behaviour signals

Mapping rationale: Rust bridge integrates task and review workflow with the development environment.

Evidence category: first_party_implementation_records; mapping confidence: medium.

Evidence gaps: Show team-wide adoption and ownership of standards before higher lifecycle level.

Klarian · klarian-02
  • Led AI and technology development for industrial pipeline monitoring and analytics, combining engineering models and machine learning to investigate pump efficiency and equipment anomalies for UK pipeline operators.
SFIA skillProvisional levelUpper candidate
Machine learning MLNG5
Scientific modelling SCMO5

Behaviour signals

Mapping rationale: Industrial cases support mixed physics/model work; personal direction is user-reported and publication confirms co-authorship and CTO context.

Evidence category: mixed_publisher_and_first_party; mapping confidence: medium.

Evidence gaps: Confirm personal model decisions, MLOps scope, team/budget and deployed operational outcomes; company work is not all personally authored.

Klarian · klarian-03
  • Co-authored published client case studies on pump-performance analysis and orifice-plate anomaly detection, connecting sensor data and physical behaviour to operational decisions.
SFIA skillProvisional levelUpper candidate
Data science DATS4
Scientific modelling SCMO4

Behaviour signals

Mapping rationale: Published pump/orifice studies connect sensor analysis and physical interpretation to decisions, with joint attribution.

Evidence category: publisher_corroborated; mapping confidence: medium.

Evidence gaps: Author contribution record and project leadership needed for level5; do not count publication as sole project ownership.

Klarian · klarian-04
  • Co-authored analysis identifying potential energy-cost reduction of approximately £5,000 per month per average mainline pump in the examined network section, conditional on restoring efficiency to as-new levels.
SFIA skillProvisional levelUpper candidate
Data analytics DAAN4

Behaviour signals

Mapping rationale: Published conditional energy-cost estimate is analysis communication, not realised benefits management.

Evidence category: publisher_corroborated_estimate; mapping confidence: medium.

Evidence gaps: Validate personal analytical contribution and measurement basis; never treat estimated £5,000/month as realised savings.

Nationwide Building Society · nationwide-03
  • Established and ran a prototyping engineering lab; developed distributed-infrastructure prototypes and contributed to AI and blockchain strategy. Sponsored and mentored the team member whose work resulted in a distributed-data-storage patent.
SFIA skillProvisional levelUpper candidate
Innovation management INOV56

Behaviour signals

Mapping rationale: Lab establishment and operation support an innovation capability; level6 is a tentative organisational-capability signal, not a global career grade.

Evidence category: mixed_institution_and_first_party; mapping confidence: medium.

Evidence gaps: Confirm organisational commitment, decision rights and innovation outcomes for6; inventor is another person.

Nationwide Building Society · nationwide-04
  • Led teams of 2–12 technical architects across parallel programmes including mobile banking, Open Banking, strategic origination, information management and regulatory change.
SFIA skillProvisional levelUpper candidate
Solution architecture ARCH5
Systems development management DLMG5

Behaviour signals

Mapping rationale: Reported architectural team span and parallel banking programmes support delivery/architecture leadership at significant work scope.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Headcount alone does not establish6; obtain mandate, delegation/resource evidence and delivery results.

Nationwide Building Society · nationwide-05
  • Set technical quality assurance and governance for supplier designs and build deliverables, including IBM and Accenture; aligned project architecture with enterprise strategy and challenged technical debt.
SFIA skillProvisional levelUpper candidate
Quality assurance QUAS5
Solution architecture ARCH56

Behaviour signals

Mapping rationale: Supplier design/build assurance fits complex-domain quality reviews and architectural conformance.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Combined with nationwide04, cross-programme governance is a candidate ARCH6 signal. Confirm target architecture ownership and functional/service/cost trade-off decisions; obtain assurance findings and organisational remit before QUAS6. No board-level GOVN inferred.

Nationwide Building Society · nationwide-06
  • Led the Synthetic Data Platform from inception to enterprise capability to support data-driven innovation and reduce reliance on live data.
SFIA skillProvisional levelUpper candidate
Solution architecture ARCH5
Systems development management DLMG5

Behaviour signals

Mapping rationale: Historical partner CV reports end-to-end platform leadership; privacy purpose supports a behaviour signal, not guaranteed privacy outcomes.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Supply delivery/adoption records, remit and outcome evidence; do not infer measured live-data reduction or ML modelling ownership.

Nationwide Building Society · nationwide-07
  • Developed privacy-preserving machine-learning and digital-twin approaches.
SFIA skillProvisional levelUpper candidate
Machine learning MLNG4
Scientific modelling SCMO4

Behaviour signals

Mapping rationale: Reported privacy ML/twin development aligns to applied modelling; sparse description limits assessment confidence.

Evidence category: historical_cv_report; mapping confidence: low.

Evidence gaps: Provide artifacts, exact role, evaluated methods and outputs; capability topics alone cannot establish5.

Nationwide Building Society · nationwide-08
  • Shaped an enterprise DevOps blueprint and demonstrated architecture as code through a proof of concept; proposed a common business-value measure for change delivery.

Behaviour signals

Mapping rationale: Enterprise blueprint and proof of concept support architecture/lifecycle practice design, with adoption explicitly unproven.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Confirm approved adoption, decision ownership and measured effectiveness before6; proposed metric is not an operated measurement capability.

M2M Ecosphere · m2m-ecosphere-02
  • Established innovation processes, an R&D ecosystem, knowledge management and organisation design to support innovation across Group Olvia.

Behaviour signals

Mapping rationale: Reported establishment of innovation processes, knowledge management and organisation design spans capability creation.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Historical CV only: seek approved designs, organisational commitment, adoption, budget/quality accountability and impact;6 is tentative INOV signal.

M2M Ecosphere · m2m-ecosphere-03
  • Deployed an AI prototype for the Group’s copper supply chain from the Democratic Republic of the Congo to China, addressing route optimisation and response to unexpected disruption.
SFIA skillProvisional levelUpper candidate
Software design SWDN4
Machine learning MLNGUngraded

Behaviour signals

Mapping rationale: Reported deployed AI prototype supports software design; AI does not necessarily mean learned models, so MLNG remains a relevant area without a level.

Evidence category: historical_cv_report; mapping confidence: low.

Evidence gaps: Confirm algorithms and personal work; prototype is not evidence of production optimisation or realised savings.

Shopitize · shopitize-02
  • Served as the most senior technology professional, aligning business strategy, product development and platform architecture. Managed up to 32 engineers and designers across four locations, then hired and built an in-house team.

Behaviour signals

Mapping rationale: Technical ownership, four-location team management and in-house transition support architecture, development and team-structure responsibilities.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: For DLMG6 the reported technology/resource remit is a candidate signal; confirm development policy, security/privacy ownership and technical/financial/quality targets. No HR-specialist RESC inferred merely from hiring.

Shopitize · shopitize-03
  • Designed and oversaw automated receipt processing using image processing and OCR; owned architecture spanning data processing, backend services, mobile APIs and a mix of public/private cloud infrastructure.
SFIA skillProvisional levelUpper candidate
Solution architecture ARCH5
Software design SWDN5

Behaviour signals

Mapping rationale: Reported ownership spans OCR processing, backend, mobile APIs and hybrid infrastructure.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Obtain architecture artifacts and decision/result evidence; do not infer ML model development from OCR alone.

Shopitize · shopitize-04
  • Designed REST APIs, publish/subscribe processing and payment integrations with BACS and PayPal; built a product-matching search engine to replace Solr.
SFIA skillProvisional levelUpper candidate
Programming/software development PROG4
Software design SWDN4

Behaviour signals

Mapping rationale: Implementation across APIs, asynchronous processing and product search supports complex software design/construction.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Show ownership across lifecycle to substantiate5; no payment compliance mandate inferred.

Shopitize · shopitize-05
  • Mentored data scientists and Science to Data Science participants.
SFIA skillProvisional levelUpper candidate
Professional development PDSVUngraded

Behaviour signals

Mapping rationale: Mentoring is direct behavioural evidence; PDSV is relevant but structured development planning and review are unshown.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Provide development plans, needs analysis and progression outcomes before assigning PDSV4 or5.

Cranfield University · cranfield-02
  • Developed wavelet-packet methods for mobile spectrum sharing and implemented the waveform on Anritsu vector signal generation and analysis hardware.
SFIA skillProvisional levelUpper candidate
Radio frequency engineering RFEN4
Formal research RSCH4

Behaviour signals

Mapping rationale: Waveform research plus instrument implementation supports radio engineering and research execution.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Confirm personal research planning, tests and external funding role before RSCH5; employment title alone gives no level.

Cranfield University · cranfield-03
  • Researched communications for in-vehicle sensor networks in an aerospace health-monitoring programme supported by an industry consortium led by Rolls-Royce and Boeing.
SFIA skillProvisional levelUpper candidate
Formal research RSCH4
Radio frequency engineering RFEN4

Behaviour signals

Mapping rationale: Sensor-communications research supports domain research/RF skill signals; consortium membership is context only.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Provide exact deliverables and contribution records; consortium names do not imply personal customer influence or full system delivery.

Cranfield University · cranfield-04
  • Developed image-processing and data-fusion methods for emitter geolocation, comparing Hough-transform approaches with particle filtering in doctoral research.
SFIA skillProvisional levelUpper candidate
Formal research RSCH4
Scientific modelling SCMO4

Behaviour signals

Mapping rationale: Institutionally deposited doctoral work supports substantive research methods, model comparison and dissemination.

Evidence category: mixed_thesis_and_first_party; mapping confidence: medium.

Evidence gaps: Confirm goal/funding authority for RSCH5; qualification and publication alone do not establish senior research management.

Selected applied AI engagements · specialist-engagements-01
  • Memrise: built a Java/Spark machine-learning pipeline for communications during a short specialist engagement.
SFIA skillProvisional levelUpper candidate
Data engineering DENG4
Machine learning MLNG4

Behaviour signals

Mapping rationale: Reported Java/Spark ML pipeline supports pipeline and ML implementation; exact operating scope is unknown.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Request source, test outcomes, personal scope and production status; no leadership level5 inferred.

Selected applied AI engagements · specialist-engagements-02
  • ThirdEye: developed a Kubernetes/TensorFlow pipeline for video recognition during a short specialist engagement.
SFIA skillProvisional levelUpper candidate
Data engineering DENG4
Machine learning MLNG4

Behaviour signals

Mapping rationale: Reported Kubernetes/TensorFlow video pipeline supports ML/data engineering implementation.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Request architecture/model evaluation and operating evidence; no accuracy or production claim.

Selected applied AI engagements · specialist-engagements-03
  • Therapy Box: developed a deep-learning speech-classification prototype for research into learning difficulties in children's speech.
SFIA skillProvisional levelUpper candidate
Machine learning MLNG4

Behaviour signals

Mapping rationale: Research speech prototype supports model development within a bounded exploratory task.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Provide model/evaluation artifacts and responsibility scope; no clinical validation or diagnostic capability inferred.

Microsharp Corporation · microsharp-01

IT/IS Executive | Microsharp Corporation | February 2003–May 2007

Designed and tested company IT and communications infrastructure with automation and redundancy.

SFIA skillProvisional levelUpper candidate
Infrastructure design IFDN4

Behaviour signals

Mapping rationale: Design and testing of company IT/communications supports infrastructure engineering.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Confirm scale, requirements, test results and design authority before5.

Technical foundation · technical-foundation-01

Platform engineering: Rust, Python, TypeScript, APIs, distributed systems, PostgreSQL, Redis, containers, CI/CD and Cloudflare Workers. Historical delivery also includes Java/Spark, TensorFlow, PyTorch and MATLAB.

SFIA skillProvisional levelUpper candidate
Programming/software development PROGUngraded
Data engineering DENGUngraded
Systems and software lifecycle engineering SLENUngraded

Behaviour signals

Mapping rationale: Technology index points to concrete project rows; tool names alone are not SFIA competence or current proficiency evidence.

Evidence category: mixed_summary; mapping confidence: medium.

Evidence gaps: Use underlying project evidence; do not assign uniform levels to every listed language/tool.

Technical foundation · technical-foundation-02

Applied AI and systems: machine learning, NLP, knowledge graphs, graph embeddings, search, image processing, sensor fusion, privacy-preserving approaches and synthetic data. Systems engineering, technical architecture, prototyping and delivery governance connect the research and implementation work.

SFIA skillProvisional levelUpper candidate
Machine learning MLNGUngraded
Scientific modelling SCMOUngraded
Solution architecture ARCHUngraded

Behaviour signals

Mapping rationale: Capability index summarises heterogeneous projects; graph embedding can be deterministic software and need not be MLNG.

Evidence category: mixed_summary; mapping confidence: medium.

Evidence gaps: Assess each project; no uniform level across all methods or recency implied.

Patents · patents-01

Secure distributed data storage | UK application GB1815423.7

Granted patent GB2574076B. Led the prototyping team and supported patenting; sponsored and mentored the named inventor.

SFIA skillProvisional levelUpper candidate
Innovation management INOV5

Behaviour signals

Mapping rationale: Team leadership and patenting support corroborate the innovation-lab story; no Alexander inventorship claimed.

Evidence category: mixed_patent_institution_first_party; mapping confidence: medium.

Evidence gaps: Duplicate support for nationwide03, not independent skill achievement; patent grant verifies legal publication, not leadership level.

Patents · patents-02

Loyalty-program technology | Patent application WO2013057491

Named co-inventor of “Method and system for providing a loyalty program” for Shopitize.

SFIA skillProvisional levelUpper candidate
Software design SWDNUngraded

Behaviour signals

  • Creativity: Named co-inventor of a technology application

Mapping rationale: Published co-inventorship is concrete creative contribution; it does not establish implementation, work autonomy or management scope.

Evidence category: patent_record_corroborated; mapping confidence: medium.

Evidence gaps: Need contribution/design artifacts and responsibility evidence to assign a software-design level.

Awards · awards-01
  • Platinum Prize, Build on Redis Hackathon, 2021: The Pattern, combining NLP-based document exploration, graph-based search and spatial VR/AR exploration.
SFIA skillProvisional levelUpper candidate
Programming/software development PROGUngraded
Software design SWDNUngraded

Behaviour signals

  • Creativity: Combines NLP, graph and spatial exploration

Mapping rationale: Organiser corroborates prize and project; recognition supports creativity without defining workplace accountability.

Evidence category: organiser_corroborated; mapping confidence: medium.

Evidence gaps: Assess implementation artifacts and contribution scope for any skill level.

Awards · awards-02
  • Best Student Paper Award, ISSPA, 2007: “Emitter Geolocation using fusion of TDOA Data with a Particle Filter”, co-authored with R. F. Ormondroyd.
SFIA skillProvisional levelUpper candidate
Formal research RSCHUngraded
Scientific modelling SCMOUngraded

Behaviour signals

Mapping rationale: Thesis reports student-paper award and co-authored modelling research; organiser certificate not inspected.

Evidence category: institutional_first_party_record; mapping confidence: medium.

Evidence gaps: Award supports research evidence, not a responsibility grade; see cranfield04.

Awards · awards-03
  • Golden Award, Codility, January 2016: Calcium2015 programming challenge.
SFIA skillProvisional levelUpper candidate
Programming/software development PROGUngraded

Behaviour signals

Mapping rationale: Personal record reports Codility recognition; no independently inspected result or workplace delivery context.

Evidence category: first_party_personal_record; mapping confidence: low.

Evidence gaps: Obtain result/certificate and solution; do not infer production proficiency or percentile.

Awards · awards-04
  • Winner, Nationwide internal IT Architecture Raspberry Pi challenge, 2017: machine-learning prototype for detecting data-compliance issues.
SFIA skillProvisional levelUpper candidate
Machine learning MLNGUngraded

Behaviour signals

Mapping rationale: Historical CVs report internal win and prototype; competition result does not establish deployed compliance capability.

Evidence category: historical_cv_report; mapping confidence: medium.

Evidence gaps: Obtain organiser/result and prototype evidence; no compliance assurance or savings inference.

Selected talks · talks-01
  • CNCF, 23 March 2023: co-speaker, “Real-time pipeline monitoring for the energy sector”.

Behaviour signals

Mapping rationale: Official event evidence supports speaking/topic attribution; audience engagement and communication impact have not been evaluated.

Evidence category: organiser_event_record; mapping confidence: medium.

Evidence gaps: No professional skill or full behavioural level from a title/listing alone; inspect recording, feedback and decision impact.

Selected talks · talks-02
  • Oxford Praxis Forum, Green Templeton College, University of Oxford, 28 November 2022: speaker, “What next after Google?”.

Behaviour signals

  • Communication: Presents specialist concepts to external audiences

Mapping rationale: Official event evidence supports speaking/topic attribution; audience engagement and communication impact have not been evaluated.

Evidence category: institution_event_listing; mapping confidence: medium.

Evidence gaps: No professional skill or full behavioural level from a title/listing alone; inspect recording, feedback and decision impact.

Selected talks · talks-03
  • EuroPython, 28 July 2021: co-presenter with Dvir Dukhan, “The Pattern: Machine Learning Natural Language Processing meets VR/AR”.

Behaviour signals

Mapping rationale: Official event evidence supports speaking/topic attribution; audience engagement and communication impact have not been evaluated.

Evidence category: organiser_event_record; mapping confidence: medium.

Evidence gaps: No professional skill or full behavioural level from a title/listing alone; inspect recording, feedback and decision impact.

Selected talks · talks-04
  • RedisConf, April 2021: “Using the Redis ecosystem to build NLP-based services”; selected in Redis’s staff-voted conference highlights for “Best I-Can’t-Believe-You’re-Using-Redis-For-THAT Story”.

Behaviour signals

  • Communication: Presents specialist concepts to external audiences

Mapping rationale: Official event evidence supports speaking/topic attribution; audience engagement and communication impact have not been evaluated.

Evidence category: organiser_event_record; mapping confidence: medium.

Evidence gaps: No professional skill or full behavioural level from a title/listing alone; inspect recording, feedback and decision impact.

Competition projects · competition-projects-01

MedGemma competition, February 2026: built a solo knowledge-grounded research prototype in Rust, including entity extraction, a knowledge graph, multi-agent orchestration, MedGemma integration and an evaluation harness; evaluated across 18 scripted cases with CPU/GPU inference.

SFIA skillProvisional levelUpper candidate
Programming/software development PROG5
Machine learning MLNG4

Behaviour signals

Mapping rationale: Solo Rust prototype and recorded 18-case harness support full prototype software responsibility and model integration, not clinical validation.

Evidence category: first_party_source_and_recorded_evaluation; mapping confidence: medium.

Evidence gaps: Recorded test evidence not rerun; integration is not proof of training novel models or production MLOps; no clinical outcome or award.

Industrial publication · publication-01

Industrial publication: Co-author, Monitoring and Anomaly Detection Approaches with AI and Data Analytics for Pipelines, Pipeline Technology Journal, 2023.

SFIA skillProvisional levelUpper candidate
Data science DATSUngraded

Behaviour signals

Mapping rationale: Publisher supports authorship and topic; underlying analytical activity is assessed in Klarian rows.

Evidence category: publisher_corroborated; mapping confidence: medium.

Evidence gaps: Do not double-count; publication alone is insufficient for DATS level.

Education · education-01

PhD | Cranfield University

Thesis: Image processing and agent-based framework for the geolocation of emitters (2010).

SFIA skillProvisional levelUpper candidate
Formal research RSCHUngraded

Behaviour signals

Mapping rationale: Institutional thesis supports research qualification and topic, not workplace responsibility level.

Evidence category: institution_corroborated; mapping confidence: medium.

Evidence gaps: Use cranfield04 research activity for indicative mapping; doctorate does not imply SFIA6/7.

Education · education-02

MSc | Bauman Moscow State Technical University

Computer Science and Computer Engineering.

Behaviour signals

Mapping rationale: University speaker biography corroborates degree subject; diploma not inspected.

Evidence category: institution_biography_corroborated; mapping confidence: medium.

Evidence gaps: No responsibility level from qualification.

Education · education-03

TOGAF 9 Certified | The Open Group

SFIA skillProvisional levelUpper candidate
Enterprise and business architecture STPLUngraded
Solution architecture ARCHUngraded

Behaviour signals

Mapping rationale: First-party TOGAF credential listing suggests framework knowledge; registry and issue details unverified.

Evidence category: first_party_credential_record; mapping confidence: medium.

Evidence gaps: TOGAF certification is not SFIA certification; assess applied architecture rows and verify registry if needed.

Updated 12 September 2026. SFIA is owned by the SFIA Foundation.