My Work

Professional experience, research, and achievements

01Experience

Jinesis AI Lab, University of Toronto·Toronto, Canada (Remote)·September 2026 – Present

→Engineered an end-to-end data-collection pipeline for the NHANES national health survey (CDC), crawling and normalizing 1,486 datasets and 49,863 variables to extend a causal-dataset-retrieval research benchmark, architected as a source-agnostic system reusable across additional data repositories
→Designed and executed a manual annotation study validating an automated paper-to-dataset linking system built on the OpenAlex API, reviewing 120 candidate pairs across 25 datasets to establish a 58% precision baseline and surface concrete failure modes
→Built and validated an LLM-based causal question-generation pipeline (OpenRouter, GPT-4o-mini) producing 201 research questions grounded in verified scientific literature, delivered as a reviewed pull request to the lab's open-source research codebase

École Polytechnique Fédérale de Lausanne, LiGHT Lab·Lausanne, Switzerland (Remote)·June 2026 – August 2026

→Selected into a research internship with a 1.2% acceptance rate at the Laboratory for Intelligent Global Health and Humanitarian Response Technologies
→Built survival and severity prediction models for Ebola Virus Disease on a 147-patient Guinea Tier 3 cohort, benchmarked against Hartley et al. (2017) and the PREDS score of Jaspard et al. (2022)

Huawei·Douala, Cameroon·August 2025 – February 2026

→Engineered Bash monitoring pipelines across distributed telecom nodes, replacing reactive recovery with automated detection and reducing downtime by over 70%
→Built a real-time fiber fault detection workflow that cut detection latency from hours to under 30 minutes
→Ran cross-layer 2G to 4G capacity diagnostics over live network telemetry, surfacing 30+ bottlenecks for resource optimisation

EleutherAI·Remote·July 2025 – August 2025

→Researched prompt optimisation strategies for reducing hallucination in large language models, evaluating prompting methods and failure modes across tasks and datasets

Engineering for Change (E4C)·New York, NY (Remote)·May 2025 – September 2025

→Built a retrieval and classification service over 8,000+ construction technology companies using Qdrant vector search, TF-IDF, and GPT-4 behind a FastAPI backend, reaching over 85% classification accuracy
→Shipped a full-stack internal research platform in Next.js and FastAPI with an LLM conversational interface, cutting analyst research time by more than 70%
→Owned the system end to end: data ingestion, embedding and indexing, API design, and the front end analysts used daily

Fatima Institute for Global AI Research·San Francisco, CA (Remote)·September 2024 – June 2025

→Investigated robustness and evaluation of multilingual foundation models across languages and cultural contexts, focusing on factual reliability and alignment
→Co-developed GlobalLies, a multilingual dataset of 170,000+ adversarial misinformation-generation prompts across 8 languages and 195 countries, quantifying disparities in LLM misinformation propagation by language and Human Development Index band
→Designed pipelines for cross-lingual consistency scoring; curated the French and Cameroonian misinformation subset (237check, Cameroon Check) covering politics, healthcare, energy, and road safety
→Co-authored "To Lie or Not to Lie?", accepted to ACL 2026 Main Conference and recognized with a Senior Area Chair (SAC) Highlight Award

→Designed and implemented a deep learning pipeline for cortical mesh segmentation that projects 3D brain surfaces into multiple 2D views, trains CNNs with extensive data augmentation, and remaps predicted segmentations back onto the 3D mesh, reaching 70% accuracy on unseen meshes
→Designed evaluation benchmarks for deformable 3D surface reconstruction using shape deviation, normal consistency, curvature preservation, and rendered appearance quality
→Benchmarked visual SLAM systems (DROID-SLAM, SAGE-SLAM, Gaussian-SLAM) in C++ for real-time 3D reconstruction on clinical endoscopy data
→Work extended into a first-author paper accepted at IEEE EMBC 2026

Meta (Facebook)·Remote·July 2024

→Evaluated 80+ early-stage African AI startups in generative AI, LLMs, and AR/VR against a rubric covering technical merit, scalability, and social impact; supported equitable selection of Meta-backed teams

Open Dreams NGO·Buea, Cameroon·September 2022 – Present

→Mentor students one-on-one through scholarship applications, essay writing, interview prep, and college selection; two mentees admitted to Hong Kong Baptist University and Ashesi University

TiC Foundation·Buea, Cameroon·March 2023 – March 2025

→Co-organized hackathons and trained 200+ secondary students across 10+ schools; led a team to 11th place nationally at the Summit competition

Cohere for AI·Remote·September 2024 – December 2024

→Developed and refined French-language prompts for Cohere for AI's Aya 8B Expanse model, improving multilingual understanding

→Reviewed AI projects across Africa, evaluating technical rigor and community impact

GDSC University of Buea / KET Foundation / Futurize·Buea, Cameroon·2023 – Present

→Core team, Google Developer Student Clubs, University of Buea (2023 – 2024)
→Student Ambassador, KET Foundation (January 2024 – Present), supporting academic outcomes for secondary students nationally
→Technical Mentor, Futurize (November 2025), supporting a team building an AI tool for farmers across Africa