Victor Ngetich

Victor Ngetich

Software & ML Engineer

UC Berkeley

I'm a full-stack software engineer specializing in turning requirements and data into fast, reliable products. At Flux Water I owned schema design and front-end polish, slashing core API latency from 2.5 s to 200 ms and lifting React/Next.js performance to 99 % on Core Web Vitals. I built AccessWASH which served thousands of users advocating for better water, sanitation, and hygiene (WASH) services across Kenya.

As a Mastercard Foundation Scholar in University of California, Berkeley’s Master of Information Management and Systems program, I'm extending that craft into machine-learning engineering. I've integrated Google Gemini and Claude APIs into 24-hour hackathon prototypes and am researching ML-powered knowledge systems that surface community data in ways policymakers can act on.

Day-to-day, I'm fluent in Python, TypeScript, React, Django/DRF, PostgreSQL, scikit-learn, pandas, Docker, and Kubernetes, with a solid grasp of system design, DevOps, and data-centric product thinking. I measure success by a single metric: how quickly real users get the information, or action, they need.

Projects

Flood Navigator

Real-time, crowdsourced flood reports (James R. Chen Award winner)
Accra pilot / Berkeley: Jan 2025 → presentRole: Backend engineer
  • Built Django + PostgreSQL API that ingests geotagged reports and clusters them on a Mapbox tile layer, powering sub-second map updates.
  • Implemented time-based push notifications and CRUD backend.
  • Led MVP launch at MIMS showcase; project now headed to Accra field pilot (Jun–Jul 2025) with local driver & market associations.
Tech stack: Django REST Framework, PostgreSQL, React PWA, Mapbox GL, Firebase Cloud Messaging, Twilio SMS

AccessWASH Web Platform

Crowd-mapped WASH coverage
Nairobi, Kenya: Feb 2023 → Jul 2024Role: Front-end Engineer
  • Architected a scalable Next.js/React front end that visualises nationwide water-access data and syncs with a micro-service back end.
  • Introduced RESTful APIs plus lazy-loading & code-splitting, boosting Core Web Vitals from 65 % to 99 % on Vercel Analytics.
  • Set up Docker-based CI/CD and drove Figma-led UX sprints that raised user retention.
Tech stack: Next.js, React, TypeScript, REST micro-services, Docker.

Lacuna

AI-powered journaling for language learners (UC Berkeley AI Hackathon 2025)
Berkeley: Jun 2025Role: Full-stack engineer, Prompt engineering
  • Built a Next.js web app that lets intermediate language learners journal daily and receive instant, private, native-level AI feedback.
  • Integrated Anthropic Claude Sonnet 4 for grammar, phrasing, and translation feedback; generated flashcards from user entries for spaced repetition study.
  • Led prompt engineering and UI prototyping with Vercel v0, iterating rapidly on user flows and feedback.
  • Supported 8 languages at launch, with smart flashcard review and calendar-based journaling.
Tech stack: Next.js, React, TailwindCSS, Anthropic Claude API, Vercel

Sugar Plum

AI women's-health assistant (Cal Hacks 11.0)
Hackathon project: Oct 2024
  • Shipped MVP in 24 h, winning shortlist mention at Cal Hacks, by pairing Next.js API routes with Google's Gemini LLM.
  • Integrated cycle-tracking prompts and evidence-based wellness tips generated on-the-fly by Gemini.
  • Styled responsive UI with TailwindCSS and deployed on Vercel for live judging.
Tech stack: Next.js, React, Gemini API, TailwindCSS.

Astronomy Discovery Portal

Celestial-object explorer
Side project: Oct 2024 → Dec 2024
  • Built a Next.js + TypeScript web app that lets users browse 100 K+ stars, galaxies, and nebulae through faceted search and instant filters.
  • Wrote fuzzy-matching algorithms for multi-field searches (name, magnitude, coordinates) to surface results in milliseconds.
  • Devised a hybrid taxonomy that bridges professional catalogues with everyday language so non-experts can explore confidently.
Tech stack: Next.js, React, TypeScript, Git.