Samwel Waweru

Software Engineer | AI Researcher

Software Engineer with over 4 years of hands-on experience building production-grade, cloud-native AI and automation systems. Experienced in designing intelligent agents, optimizing model performance, and integrating automation workflows at scale. Published ML researcher with international conference presentations and a strong engineering foundation that bridges AI, systems design, and real-world product delivery.
Now that that part is done 🥱 Press the Start button below to continue. Yes, the green one. I'm doing Windows XP things here.

kamausamwelwaweru@gmail.com

github.com/samkamau81

linkedin.com/in/samuelwaweru

New York City, USA

orcid.org/0009-0005-7859-5408

EDUCATION

M.S. Artificial Intelligence (Robotics and Perception Concentration)

Columbia University, New York, USA

Research focus on on-device intelligence, efficient inference, multimodal AI systems, computer vision, robot learning, retrieval systems, and embedded machine learning under resource-constrained environments
Relevant Coursework: Machine Learning, Computer Vision I & II, Embedded AI, Natural Language Processing, Robot Learning, Optimization for Intelligent Systems

B.Sc. Mechatronic Engineering, Robotics

Jomo Kenyatta University of Agriculture and Technology, Nairobi, Kenya

Capstone Project (Rank 1/18): Delta Laser-Weeding Robot with computer vision and YOLOv11
Coursework: Calculus 1-4, Advanced Robotics, Differential Equations, Computer Science

PROFESSIONAL EXPERIENCE

AI Expert

EdTech Hub, Result for Development, Nairobi, Kenya | Oct 2025 – Present

Engineered AI-powered backend services for education platforms at national scale
Led design and deployment of secure LLM integrations ensuring compliance with governance and data protection
Built and validated multiple proof-of-concept AI services with optimized model pipelines

Machine Learning Engineer (Kaggle BIPOC)

Kaggle USA, California (Remote) | Dec 2022 – Oct 2025

Developed production-grade LLM integrations including RAG-powered analytics engine using FAISS and Pinecone
Built AI agent development platform enabling secure deployment of multi-agent LLM workflows to embedded IoT systems
Fine-tuned Gemma-7B using LoRA, PEFT, and quantization for cybersecurity Q&A
Collaborated with globally distributed engineering teams on LLMOps and secure model integration

RESEARCH & PUBLICATION

Edge-Optimized YOLOv11-Based Real-Time Weed Detection for Autonomous Precision Agriculture

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2025) | Hangzhou, China (Oral Spotlight + Poster Presentation)

Authors: Samwel Waweru (Lead Author), Ray Njoroge

Novel approach to real-time weed detection using optimized YOLOv11 models on resource-constrained edge devices for precision agriculture applications.

View Paper →

Biodiversity Conservation Protocol in Meru County for Fire Prevention using Drone Technology

Forum for Global Challenges, University of Birmingham | May 2022 (Oral)

Authors: Jackson Mutembei, Samwel Waweru (Contributor)

Drone-based monitoring system for early fire detection and biodiversity conservation in Kenyan wildlife areas.

View Paper →

Projects

ENGINEERING PROJECTS

Golden Record (National Data Integrity System)

Solution:
Designed an AI-powered Golden Record system to resolve duplicate and inconsistent learner and school records in NEMIS, enabling a single trusted source of truth for education planning, funding, and policy decisions.

Engineering Work:
Built a data ingestion and normalization pipeline integrating learner, school, attendance, and bursary records. Implemented rule-based entity resolution as a baseline, followed by AI similarity scoring and human-in-the-loop validation workflows. Designed flagging mechanisms for under-reporting, over-reporting, and data quality gaps, with audit trails for governance and transparency.

Tech Stack:
Python, Entity Resolution, Fuzzy Matching, Data Pipelines, Human-in-the-Loop Systems, Synthetic Data Generation, Analytics Dashboards

EdTech Hub Project: NDIS

FinGPT (Financial Language Model System)

Solution:
Implemented a backend service for a domain-specific financial LLM, providing structured, reliable real-time inference with strong schema guarantees and backward-compatible APIs.

Engineering Work:
Designed FastAPI services with well-typed Pydantic models, implemented input validation, rate management, and caching, and integrated financial text-processing pipelines with maintainable REST endpoints.

Tech Stack:
FastAPI, Pydantic, Python, REST API Design, Async I/O, Financial Text Processing

GitHub: Fin GPT

Delta Laser Weeding Robot (Computer Vision & Robotics Backend)

Solution:
Built a low-latency computer vision inference service for autonomous weed detection, enabling precise robotic laser control in field conditions.

Engineering Work:
Optimized YOLOv11 inference on Raspberry Pi 4 using INT8 quantization and I/O pipeline tuning. Designed structured interfaces exposing detections to motion controllers and debugged data flow between the CV service, G-code generator, and microcontroller firmware.

Tech Stack:
Python, YOLOv11, TensorFlow, OpenCV, Raspberry Pi, Microcontroller Integration

GitHub: Delta Robot

EmbedMind Studio (LLM-Powered Embedded Agents)

Solution:
Developed a lightweight platform for assembling, simulating, and deploying LLM-powered agents on embedded and edge devices.

Engineering Work:
Built a Python backend orchestrating agent logic and device communication over serial/WiFi, integrated Gemini and OpenAI APIs, containerized development environments with Docker, and implemented CI/CD workflows for reproducible builds and deployments.

Tech Stack:
Python, C++, Docker, CI/CD Workflows, Gemini/OpenAI APIs, Embedded Networking (ESP8266, Raspberry Pi)

GitHub: EmbedMind Studio

MentalMuse (NLP Mental Health Analysis)

Solution:
Created a secure Natural Language Processing analysis pipeline for mental-health text, focused on stable, explainable processing and safe data handling.

Engineering Work:
Implemented scalable Python services and orchestrated batch and streaming workloads using Snowpark. Built secure API integrations routing users to support services and added lightweight explainability around NLP outputs.

Tech Stack:
Python, NLP Pipelines, Snowpark (Snowflake), Secure API Integration, Data Validation

GitHub: Mental Muse

Kilimo-AI (LLM-Assisted Smart Irrigation)

Solution:
Built an LLM-assisted irrigation control system that translates natural-language input into reliable device-level commands for low-power IoT hardware.

Engineering Work:
Designed Python–C++ modules for prompt processing and decision logic, integrated cloud REST APIs for device state updates and alerts, and optimized system behavior for low-power rural deployment environments.

Tech Stack:
Python, C++, LLM Prompt Engineering, IoT Controllers, REST APIs, Cloud Integration

GitHub: Kilimo AI

My Pictures