Worcester, Massachusetts · Data Scientist · AI Engineer · Agentic AI Developer · Engineering Analyst

Tafirenyika
Shoniwa

Data Scientist · AI Engineer · Agentic AI Developer · Engineering Analyst

“I build and study AI systems people can trust.”

Based in Worcester, Massachusetts, I work across data science, AI engineering and systems analysis: Bayesian inference and forecasting research (Bayesian cyber risk, disease signal detection, overdose mortality), and production AI systems I designed and built end to end, Nexa, Shoviqo and Shoviqo Dispatch. Every title above is backed by the projects on this page, and my education is set out separately, qualification by qualification, in the Education section below.

On LinkedIn: linkedin.com/in/tafirenyika-shoniwa-ts1009 · On GitHub: github.com/Tafirenyika-AI

Data Science & Bayesian InferenceAI Engineering Agentic AIEngineering Analysis
Tafirenyika Shoniwa presenting at a lectern at Clark University, wearing a dark suit and a Clark University lanyard
Tafirenyika Shoniwa, presenting at Clark UniversitySpeaking at a Clark University Graduate Student Council event, leadership and public speaking alongside the research and engineering work below.
71,000+ itemsNexa knowledge corpus
4 live AI productsNexa · Studio · Shoviqo · Dispatch
BCommInformation Systems degree
7Public repositories on GitHub: open to read now
4Live AI products built & shipped
71,000+Items in the Nexa knowledge corpus
What I Do

Four titles, each one shown by the work on this page

Listed in the order of the evidence behind them. Every card links to the projects and credentials that prove it, so you can check the claim against the work itself.

01 · Strongest research evidence

Data Scientist

Statistical modelling, Bayesian inference and forecasting on real health, security and public-policy data, with public, reproducible code and honestly scored predictions.

02 · Production systems, end to end

AI Engineer

I design, build and ship complete production AI systems, models, serving, apps, backends and operations, not demos that stop at a notebook.

Shown by
04 · Systems & process analysis

Engineering Analyst

Information-systems analysis of complex operations: freight and logistics workflows, routing and navigation systems, compliance gates and business intelligence.

Shown by
About

Data Scientist · AI Engineer · Agentic AI Developer · Engineering Analyst

A professional summary for employers, in the same order as the work itself: data science first, then AI engineering, agentic systems, and engineering analysis.

Professional profile

I'm Tafirenyika Shoniwa: a Data Scientist, AI Engineer, Agentic AI Developer and Engineering Analyst based in Worcester, Massachusetts.

Undergraduate education: I hold a BComm in Information Systems: a Bachelor of Commerce focused on information systems, business intelligence and data analytics.

Graduate education: separately, my graduate study is in Data Analytics (the MSDA programme) at Clark University in Worcester, including its capstone research. The two qualifications are listed separately in the Education section.

As a data scientist, my research covers Bayesian cyber risk quantification, cancer malignancy prediction (Silent Signals), early disease signal detection, air pollution and respiratory health, and county-level overdose mortality forecasting, all of it public and reproducible on GitHub. As an AI engineer, I build production systems end to end: Nexa (research intelligence on its own trained brain), Shoviqo (mobile navigation and computer vision), Shoviqo Dispatch (freight operations and compliance), and an AI audio & speech platform with a structured speech-data pipeline. As an agentic AI developer, those systems perceive, reason, plan and act in governed loops; as an engineering analyst, I analyse the systems and processes underneath them, logistics workflows, routing, compliance and business intelligence.

I also present and lead in person, including speaking at a Clark University Graduate Student Council event (pictured above). Across all of this work I emphasise reproducible analysis, clear documentation, and measured evaluation, results that can be checked, not just claimed.

“The world no longer rewards claiming to code line by line. It rewards knowing structure, tools and systems, and innovating.”
My working philosophy: portfolio and shipped systems over claims

At a glance

  • 📍Based: Worcester, Massachusetts, USA
  • 🎓Undergraduate degree: BComm in Information Systems, see Education
  • 🎓Graduate study: Clark University, Worcester, MA, Data Analytics (MSDA programme), see Education
  • 📊Focus: Data Science · AI Engineering · Agentic AI · Engineering Analysis
  • inLinkedIn: linkedin.com/in/tafirenyika-shoniwa-ts1009 ↗ · 1K followers · 500+ connections
  • ⌥GitHub: github.com/Tafirenyika-AI ↗ · 7 public repos · private projects on request
How I Build · Agentic AI

One agentic loop, applied to everything

This is the Perceive → Reason & Plan → Act → Learn cycle behind Nexa's autonomous research workflow, and the working method behind every system I ship.

1

Perceive: start from real data

I gather the actual signals first: user problems, datasets, road data, market data, research literature. Nexa's 71,000+ item corpus and Shoviqo's own recorded drives both started here.

2

Reason & Plan: structure before code

Feasibility first, architecture second. I map the system, its safety and legality gates, and how success will be measured, before building anything.

3

Act / Build: ship the whole product

Models, apps, backends, documents, payments logic: I build end to end and put live products in front of real users, not demos that stop at a slide.

4

Learn & Improve: let evidence decide

Predictions get Brier-scored, model gates get measured, failed verdicts get published honestly. What the evidence kills, I stop. What it proves, I scale.

Projects

Projects & research, with real thumbnails and real links

Ordered to match my titles: data science & research first, then AI engineering products, then agentic & trading AI. Public projects show their real GitHub card and link straight to the repository. Live products show a real screenshot of the live site and link to it. Private code is marked Private: Access on Request.

1 · Data Science & Research: public on GitHub

The evidence behind Data Scientist. Each thumbnail is the project's real GitHub repository card, click any thumbnail or title to open the code, data and notebooks.

GitHub repository card for Tafirenyika-AI/-Bayesian-Cyber-Risk-Quantification-System
Data Science · Bayesian Inference · Public

Bayesian Cyber Risk Quantification System

A Streamlit-powered web application that uses Bayesian inference with MCMC simulation to quantify the financial impact of cyber risk, turning uncertain threat signals into decision-ready numbers. Python · Streamlit · Bayesian MCMC.

View on GitHub ↗● Public repo
GitHub repository card for Silent Signals, Environmental and Socioeconomic Drivers in Cancer Malignancy Prediction
Data Science · Health ML · Public

Silent Signals: Cancer Malignancy Prediction

Environmental & socioeconomic drivers in cancer malignancy prediction, exploring the weak, overlooked signals in health data that only show up when you model carefully. Jupyter Notebook research.

View on GitHub ↗● Public repo
GitHub repository card for The Impact of Air Pollution on Respiratory Health and Economic Costs
Data Science · Health Informatics · Public

Air Pollution, Respiratory Health & Economic Costs

A multi-sector analysis connecting air pollution exposure to respiratory health outcomes and economic costs, modelling showed a clear correlation between pollution and spending, with visualisations confirming the trend patterns. Python.

View on GitHub ↗● Public repo

Also in this research family: MSDA Capstone, county-level overdose mortality risk forecasting (spatiotemporal, interpretable ML, with Frederick Baafi & Eddington Shoko), MOSAIC DRI and AgroZim: detailed in Research below and on my LinkedIn profile ↗.

2 · AI Engineering: live products I built end to end

The evidence behind AI Engineer and Engineering Analyst. Each thumbnail is a real screenshot of the live product, click through to use it. The code repositories are private; access is available on request.

Real screenshot of the live Nexa sign-in page at tashiq-nexa.vercel.app Live screenshot: tashiq-nexa.vercel.app (sign-in)

NEXA

Autonomous research intelligence running with no commercial LLM APIs: its own trained brain over a 71,000+ item knowledge corpus, working the agentic loop: Perceive → Reason & Plan → Act → Learn. Includes Schoolhouse (preschool to PhD, exam-gated learning), NexaG virtual-GPU serving (CPU pool router, measured 39.4 tok/s two-node proof), a Brier-scored Prediction Ledger, and PDF / Word / Excel document generation, governance-first by design.

AI EngineeringAgentic AILLM TrainingLoRARAGGovernance
Real artwork from the live site: studio.tashiq.com

AI AUDIO & SPEECH PLATFORM

Speech and audio AI system (repository: tashiq-studio, TypeScript): multitrack recording workflows, audio enhancement and processing pipelines, and Shona Foundry: a structured speech-data collection system with a 180-unit phonetic inventory, a 10-level contributor recorder, tiered consent, and quality review for dataset building and model training.

Speech AIAudio ProcessingDataset EngineeringTypeScript
Real screenshot of the live Shoviqo admin console sign-in at shoviqo.vercel.app Live screenshot: Shoviqo web admin console, shoviqo.vercel.app/admin

SHOVIQO

Phone-first AI driving copilot for iOS & Android. Navigation engineering on MapLibre + OpenStreetMap + Valhalla with live traffic, a 3D realistic view, satellite layer, camera-based lane awareness and static sign reading, dashcam recording, and trip history. Travel modes cover car, bus, train, bike, walking, flight, and rideshare. Launched in Worcester, MA: region is configuration, not code, so it is built to work anywhere.

FlutterComputer VisionGIS & RoutingMobile
Real screenshot of the live Shoviqo Dispatch sign-in page at shoviqo-dispatch.vercel.app Live screenshot: shoviqo-dispatch.vercel.app (sign-in)

SHOVIQO DISPATCH

AI-powered global freight operations console, systems analysis of a full logistics operation: org-scoped shipments, real PDF documents (BOL, rate confirmation, invoice, inspection report), FMCSA compliance, and the BorderReady legality gate with regional rule packs for the US, EU, South Africa, and SADC. Its AI copilot and deterministic automation engine act on live records; it runs its own double-entry accounting engine, a members-only marketplace, and 6 languages.

Agentic AILogistics SystemsComplianceAccounting EngineNext.js

3 · Agentic AI: prediction & trading agents

The evidence behind Agentic AI Developer: autonomous agents that forecast, act and get scored on real outcomes.

GitHub repository card for Tafirenyika-AI/AI-Forex-Signal-Desk: Live decisions, real outcomes
Agentic AI · Forecasting · Public + Private

Apex, Prediction Systems & AI-Forex-Signal-Desk

Autonomous trading and forecasting research: a Deriv auto-trader with an overnight sweep (demo), a prediction-market engine (Kalshi temperature markets, Phase 0 shadow scanner over 876 settled events), a Forward Prediction Ledger scored with Brier scores, and the public AI-Forex-Signal-Desk (Python: “Live decisions, real outcomes”). Verdicts are published honestly, including the failures.

Engineering Analysis · Capstone

MSDA Capstone: Overdose Risk Forecasting

County-level drug overdose mortality risk forecasting: a spatiotemporal, interpretable machine-learning study built with Frederick Baafi and Eddington Shoko as the MSDA 3999 capstone team at Clark University, data science and engineering analysis applied to a public-health system.

Spatiotemporal MLPublic HealthInterpretability
GitHub · @Tafirenyika-AI

Projects on GitHub: categorised, with real links

Categorised in the same order as my titles: data science & research (public, open the code now), AI products & platforms (live products, private code), and trading & finance AI. Every thumbnail is the project's real GitHub card or a real screenshot of the live product.

TS

Tafirenyika-AI on GitHub

github.com/Tafirenyika-AI, public research repositories are open to read right now. Product repositories are private: use the live links on each card to see the product working, and request code access on LinkedIn.

🔒 Some of my GitHub projects are private. Access is available on request, contact me on LinkedIn and I will grant access for genuine professional review. Public repositories and live products below are open to everyone, no request needed.

Request Access on LinkedIn ↗

A: Data Science, Health & Research

All public, real GitHub cards, click through to the code

Also in this research family: MSDA Capstone: Overdose Mortality Risk Forecasting (spatiotemporal, interpretable ML with Frederick Baafi & Eddington Shoko), MOSAIC DRI and AgroZim: see Research below and my LinkedIn profile ↗ for the full record.

B: AI Products & Platforms

Live products · Code private, access on request
Real screenshot of the live Shoviqo admin console sign-in
TypeScript · Dart · Python · JavaScript🔒 Private · 4 repos: Access on Request

Shoviqo: Driving Copilot (4 repositories)

Phone-first driving copilot for iOS & Android, navigation, 3D realistic view, lane awareness, sign reading and dashcam. Four repos grouped as one product:

  • shoviqo: TypeScript · Private
  • shoviqo-app: Dart / Flutter · Private
  • shoviqo-routing: Python · Private
  • shoviqo-preview: JavaScript · Private
Python🔒 Private: Access on Request

Matemai

Repository: Matemai · Language: Python · Visibility: Private. Details and code access are available on request.

C: Trading & Finance AI

Forecasting agents, signals & automated trading research
Trading System🔒 Private: Access on Request

apex-ai-trading

Apex AI trading system, Deriv auto-trader, overnight sweep and prediction-market research, run on demo with honest, measured verdicts including failures.

See everything in one place

Full project history, credentials and experience are on LinkedIn. All public code is on GitHub. Private products are best seen live, every live link above opens the real product, and private code access is available on request.

Skills

Skills: grouped to match my titles, in the same order

Data science first, then AI engineering, agentic AI, engineering analysis, and the software engineering underneath all of it.

📊

Data Science & Analytics

Data AnalyticsBayesian InferenceBayesian Risk QuantificationStatistical ModellingForecastingSpatiotemporal MLHealth InformaticsData StorytellingGoogle Analytics
🧠

AI Engineering

Production AI SystemsLLM Training & LoRA Fine-tuningModel Serving (NexaG)RAG / Corpus BuildingComputer Vision: Lane & Sign DetectionSpeech & Audio AIAI Evaluation & Governance
🤖

Agentic AI Development

Autonomous AgentsMulti-step Reasoning WorkflowsPerceive → Reason → Act → Learn LoopsAI Copilots on Live RecordsDeterministic Automation EnginesPrediction Systems & Brier Scoring
🧭

Engineering Analysis & Systems

Information Systems AnalysisBusiness IntelligenceFreight & Logistics SystemsRouting & Navigation EngineeringCompliance & Legality GatesSystems / Process AnalysisTechnical Documentation
🛠️

Software Engineering

PythonTypeScript / Next.jsFlutter / DartFastAPIPostgreSQL / NeonVercelGoogle Cloud RunGitHubMapLibre · OpenStreetMap · ValhallaSystem Architecture
✦

Professional

Data StorytellingPublic Speaking & PresentingTechnical DocumentationProduct Design & UXResearch Communication
Education

Education: two separate qualifications, listed separately

Structured the way my LinkedIn profile presents it: my graduate study and my undergraduate degree are two different qualifications at two different levels: they are never combined into one line or one credential.

Graduate Study

Clark University

Graduate study in Data Analytics, MSDA programme

📍 Worcester, Massachusetts, USA

My current graduate university in the United States, where my study and research sit in data analytics, AI systems and health informatics.

Evidence at Clark
  • MSDA capstone: spatiotemporal, interpretable ML forecasting of county-level drug overdose mortality risk, with Frederick Baafi and Eddington Shoko. See Research.
  • Graduate Student Council: presented “Bold Statements: A Moment of Leadership”; the photo in the hero above is from this event.
  • Research focus: AI systems, data analytics and health informatics.
Undergraduate Degree

BComm in Information Systems

Bachelor of Commerce (BComm), Information Systems

My undergraduate degree, the foundation of my data science, business intelligence and engineering-analysis work: how organisations run on information systems, data and process.

Focus areas
  • Information Systems
  • Business Intelligence & Data Analytics
  • AI & Machine Learning
  • Data Science
Information SystemsBusiness Intelligence & Data AnalyticsAIMachine LearningData Science
Licenses & Certifications

Licenses & certifications

Listed separately from my education, exactly as on my LinkedIn profile, each with its issuer and date.

MAR 2025

Health Informatics

Johns Hopkins University School of Medicine

MAY 2020

CompTIA Security+

CompTIA

2023

Data & Analytics Certificates

Corporate Finance Institute (CFI)

AUG 2023

Advanced Google Analytics

Google

APR 2023

Deloitte STEM Technology Virtual Experience Program

Deloitte

Memberships

Memberships

📊

Member, International Society for Data Science and Analytics (ISDSA)

Oct 2025 to Present

Verify on LinkedIn ↗
🏥

Member, American Public Health Association (APHA)

Public health membership alongside my health informatics and public health research, including early disease signal detection and air pollution and respiratory health.

🎓

Member, Honor Society

An organisation recognising academic achievement and leadership.

honorsociety.org ↗
Research

Research & capstone work: the Data Scientist evidence

Research & capstone

MSDA Capstone: Overdose Mortality Risk

Spatiotemporal, interpretable ML forecasting of county-level drug overdose mortality risk. Team: Tafirenyika Shoniwa, Frederick Baafi, Eddington Shoko.

MOSAIC DRI · Silent Signals

Applied research projects on data systems and weak-signal detection, listed on my LinkedIn research portfolio. Silent Signals code: View on GitHub ↗

Bayesian Cyber Risk Quantification System

Probabilistic risk numbers for security decisions, statistics meeting CompTIA Security foundations. View on GitHub ↗

AgroZim · Air Pollution & Respiratory Health

Zimbabwe agriculture innovation, and health-informatics research linking air quality to respiratory outcomes and costs. Air Pollution on GitHub ↗

Early Disease Signal Detection: U.S. Communities

Interpretable multi-source AI framework for early disease signal detection in U.S. communities. View on GitHub ↗

Prediction Ledger

A public-discipline forecasting record with Brier scoring, 53 forecasts entered exactly as written, settled openly as they resolve.

Research interests

Bayesian methods for risk and decision-making, interpretable machine learning for health and public-health questions, early signal detection in noisy multi-source data, and rigorous evaluation of AI systems, including scored forecasting and transparent reporting of results.

Full details, credentials and project history are on LinkedIn and public code is on GitHub (@Tafirenyika-AI). Private project code is available on request via LinkedIn.

Open to professional opportunities

For roles and professional enquiries in data science, AI engineering, agentic AI development and engineering analysis: and for access to my private GitHub projects, please contact me via LinkedIn. My website is tashiq.com; LinkedIn is the best place to verify my credentials and experience.

LinkedIn: linkedin.com/in/tafirenyika-shoniwa-ts1009 · GitHub: github.com/Tafirenyika-AI