Skip to main content

Hi, I'm

Sam

AI/ML | Cryptography

Most engineers build locks. Some build lockpicks. I study the hands that hold them.

Full-cycle developer with experience shipping intelligent autonomous and decentralized systems. I enjoy building solutions that make everyday tasks and processes easier. I also like reading systems synthesis and convergent engineering research.

AI/Data engineer, Vodacom. Previously AI Agentic Engineer intern at Nethermind where I contributed to Agentic and ZK engineering initiatives.

Skills

AI/ML

  • Deep Learning
  • Computer Vision
  • Language Models, RAG
  • Generative AI
  • Agentic Systems/Orchestration

Stack

  • Python
  • TypeScript
  • SQL
  • Rust

Platforms & Infrastructure

  • Vector DBs
  • Graph Memory
  • AWS
  • GCP
  • Cloudflare

Blockchain Development

  • Onchain Data Extraction
  • Ecosystem/Network Research
  • DeFi Analysis
  • Hardhat

Data Operations

  • Data Engineering
  • RPA
  • Digital Transformation
  • Solution Delivery

Experience

AI Agentic Engineer Intern

Nethermind

Jan 2025 – Apr 2025

London, UK

  • Engineered autonomous agents with LLM-driven reasoning cycles, modular toolchains, and orchestration ADKs.
  • Collaborated with ZK engineering research teams to prototype agent orchestration strategies across distributed environments.

Blockchain Data Analyst

Securities.io

Jun 2022 – Sep 2023

San Francisco, CA

  • Led quantitative research on Layer 1/2 networks, synthesizing reports on gas economics, consensus, and crosschain activity.
  • Authored technical publications on zk-proof systems, tokenomics, smart contract security, DeFi exploits and compliance.

Cloud/ML Trainee

Azubi Africa (AWS Re/Start)

Jan 2022 – Apr 2022

Nairobi, KE

  • Designed and provisioned cloud-native analytics pipelines and app stacks using AWS architecture best practices.

IoT Systems Engineer Attachee

Gearbox

Feb 2021 – May 2021

Nairobi, KE

  • Prototyped motion-sensing IoT systems using Raspberry Pi/Arduino; integrated alert logic and edge processing.
  • Validated sensor network designs via circuit simulation and proof-of-concept field testing.
  • Configured Linux-based IoT gateways to bridge PLCs with telemetry and monitoring pipelines.

IT Support Attachee

Ministry of Information, Communications and Digital Economy

Jul 2019 – Oct 2019

Nairobi, KE

  • Automated system patching workflows using WSUS and MECM (formerly SCCM).
  • Diagnosed network disruptions (DNS, SMTP) via Wireshark captures and traceroute-based packet path analysis.
  • Deployed Cisco and Juniper network infrastructure; configured firewall rules, VLANs, and switch logic.
  • Resolved application faults through log tracing and dynamic hotfix deployment strategies.

Projects

Q&A Retrieval Augmented Generation

Quetzal is an end-to-end RAG system for information retrieval and question answering. By integrating web crawling, document processing, vector storage, and generative AI, the system provides accurate, source-attributed answers to complex research queries.

Gemini Weaviate Flask Python
NDA

AI Agentic Workflow Automation Suite

AI agent management platform leveraging Next.js, TypeScript, and FastAPI. The platform orchestrates autonomous AI agents with workflow automation capabilities, integrating with Google Workspace, Slack, Notion, Telegram, GitHub, and LangFuse for observability.

Next.js TypeScript FastAPI Redis MongoDB Okta LangFuse
NDA

Multilingual ASR Fine-tuning

Fine-tuned OpenAI Whisper to optimize recognition of English, Swahili, and Sheng in complex, code-switching conversational contexts. Built a robust hybrid pipeline for the automated segmentation, normalization, and processing of diverse datasets from Mozilla Common Voice and HuggingFace, leveraging LLM-augmented analysis to ensure high-precision transcription accuracy across dialectal variations.

Whisper HuggingFace PyTorch Python

Zero-Knowledge ASP System

A zero-knowledge attestation service provider PoC system enabling private and compliant blockchain transactions. Combines optimized Circom circuits using PLONK proving schemes with a multi-chain compatible architecture to verify Merkle tree-based membership claims.

Circom PLONK Solidity Python Merkle Trees

Reads

Reading List

  • Thinking in Bets Annie Duke
  • The Master Algorithm Pedro Domingos
  • Hello World Hannah Fry
  • Superforecasting P. E. Tetlock & Dan Gardner
  • The Anthropic Cosmological Principle John D. Barrow
  • The Demon in the Machine Paul Davies
  • Computational Logic and Human Thinking Robert Kowalski
  • The Model Thinker Scott E. Page
  • The Alignment Problem Brian Christian
  • How to Expect the Unexpected Kit Yates

Connect

Reach out if you're working on exciting projects in AI or cryptography, or wish to discuss potential collaborations.