$ whoami
Francis
O’Hara Aidoo▍
I build full-stack software that makes AI useful, from frontend and backend to the pipelines underneath.
CS @ Colby College '27 · 2x Google SWE Intern · Platform AI @ Medidata · Co-founder @ StyleSyncs
~/profile.jpg
now: Platform AI @ Medidata
ARR unblocked at Google
Officially published Terraform resource for Vertex AI
users & $9K raised
StyleSyncs, the startup I co-founded
CSV records/hour
Apache Beam transform on Google Cloud Dataflow
CS courses TA'd
From intro programming to Neural Networks at Colby
00 · cat about.md
I'm a Computer Science student at Colby College ('27, AI concentration) who builds full-stack software with a deep end in AI and cloud infrastructure.
At Google, I shipped an officially published Terraform resource for deploying generative AI models on Vertex AI: public, documented, in production for enterprises today, and the key that unblocked a $19.2M ARR opportunity for Google Cloud partners. The summer before, I built an Apache Beam transform that processes 3B+ CSV records an hour on Google Cloud Dataflow. Right now I'm on the Platform AI team at Medidata, building agentic LLM pipelines with frontier models on Amazon Bedrock that extract structured data from clinical trial documents, for products like Protocol Digitizer and eConsent.
Across all of it, what motivates me most is shipping things people actually use, whether that's frontend, backend, or the pipelines underneath.
01 · experience.log
The pipeline so far
What I shipped at each stop: two Google internships, an AI platform for clinical trials, a startup I co-founded, and the classroom where it all started.
Medidata Solutions
stage: extractData Science Intern · Platform AI
May 2026 → Present · New York, NY
Building LLM pipelines with frontier models on Amazon Bedrock for a platform powering 37,000+ clinical trials worldwide.
- Engineering an agentic LLM pipeline for Medidata's Protocol Digitizer service, automating the parsing of Schedule of Assessments tables, footnotes, and procedural metadata from clinical trial protocols.
- Improving the accuracy of the LLM extraction service behind Medidata eConsent through prompt engineering, schema validation, model comparison, and ground-truth evaluation of each release.
PythonAmazon BedrockClaudeLLMOpsNLPAmazon S3
Google
stage: deploySoftware Engineering Intern · Vertex AI
May 2025 → Aug 2025 · Sunnyvale, CA
Shipped public infrastructure for deploying generative AI models at scale on Google Cloud.
- Engineered an officially published Terraform resource in Go for deploying generative AI models on Vertex AI, unblocking a $19.2M ARR opportunity for Google Cloud partners.
- Discovered and resolved a critical race condition in the Vertex AI backend (Java) that blocked concurrent model deployments; validated the fix with JUnit and Mockito test suites.
- Built a chat-based multi-agent system with Google's Agent Development Kit whose sub-agents autonomously discover open-source GenAI models, provision endpoints, deploy models, and run inference.
Google
stage: transformSoftware Engineering Intern · Apache Beam
May 2024 → Aug 2024 · Seattle, WA
Built high-throughput data infrastructure for migrating legacy data into the cloud.
- Engineered a schema-aware Apache Beam transform in Java processing 3 billion+ CSV records/hour on Google Cloud Dataflow.
- Ensured data integrity and fault tolerance by validating schema constraints ahead of execution and routing malformed records to a Dead Letter Queue.
JavaApache Beam
Cloud Dataflow
Terraform
StyleSyncs
stage: generateCo-Founder & Engineer
Jan 2026 → Present · Waterville, ME
Co-founded an AI-powered fashion visualization platform with 100+ registered users and $9K+ raised through pitch competitions.
- Developing the platform's flagship real-time video try-on feature that overlays clothes on live camera feeds.
- Engineering a generative pipeline based on Gemini 3 Pro Image to power the platform's static virtual try-on and digital avatar creation features.
Colby College
stage: teachUndergraduate Teaching Assistant · CS
Sep 2024 → Present · Waterville, ME
Selected by faculty to mentor students across 10+ CS courses, from intro programming to Neural Networks and Computer Organization.
- Support a cohort of 100+ students in courses like Data Structures & Algorithms, running code reviews and debugging sessions in Python, Java, C, and VHDL.
PythonJava
CVHDLNeural Networks
02 · ls ~/work
Selected work
Four builds I'm proudest of, each shipped end to end. Click through for the full case study.
also shipped
Vision Assistant
Offline-capable OCR device on Raspberry Pi for visually impaired users. Hybrid Cloud Vision / Tesseract pipeline with text-to-speech and ultrasonic proximity feedback.
Multi-Agent Model Deployer
Chat-based multi-agent system built at Google with the Agent Development Kit: sub-agents discover open-source GenAI models, provision endpoints, deploy, and run inference autonomously.
03 · tail -f roadmap.log
What I'm shipping next
A few things I'm actively building to deepen my expertise. Current projects span distributed systems, mobile development, and harness engineering.
Mule Mart 2.0: Go microservices
in progressRebuilding Mule Mart's backend as event-driven Go microservices ahead of a public launch: gRPC for service-to-service calls, a GraphQL gateway to aggregate reads across services, Kafka for buyer-seller messaging and notifications, PostgreSQL per service, all deployed to Kubernetes with Prometheus and Grafana observability baked in.
Go
gRPC
GraphQLKafka
PostgreSQL
KubernetesPrometheusGrafanaPocket agent: on-device AI for iOS
up nextA native Swift iOS app running open-weight models like Gemma 4 E4B fully on-device with MLX, tool-calling into Calendar, Reminders, and Shortcuts. Wakes on background triggers to check things and notify you proactively. No server, no API key, works in airplane mode.
SwiftSwiftUImlx-swiftEventKit
App IntentsBackgroundTasksAgent harness in Rust
queuedA model-agnostic agentic coding harness in Rust that runs any open-weight model, with an agent loop, tool runtime, context management, sandboxed execution, and a ratatui-based terminal UI. Validated with a custom Python eval suite, then benchmarked on a subset of SWE-bench Verified.
DockerPython04 · skills --list
Tools I ship with
Languages, frameworks, and infrastructure that have shipped to production somewhere. Click through the tabs.
~/francis-ohara/skills
❯ skills --group languages
# compiled, scripted, and everything between
PythonGoSwift
TypeScript
JavaScriptC/C++# always_learning: true
05 · git log --author=francis
Shipping in public
A live look at what I've been pushing, pulled straight from the GitHub API at build time.
1,202 contributions in the last year
refreshed hourly
1,202
contributions
past year
current streak
Jul 27 – Aug 8
22
longest streak
Dec 31 – Jan 21
17
public repos
34
followers
176
days active in the last year
2019
shipping on GitHub since
06 · ./connect.sh
Let’s build something
worth shipping.▍
Open to 2027 new-grad SWE & data science roles. If you’re hiring engineers who care about AI, infrastructure, and products people actually use, I’d love to talk.