Computer Engineering + Business · Boston University ’27
I build AI products for decisions people need to trust. My work connects model reasoning with verifiable evidence, constraints enforced in code, and human judgment—from understanding student work to governing what an agent can do.
LinkedIn · Email · Asili · Featured by BU Engineering
Founder, Asili — Building evidence for consequential decisions, starting in education. As AI changes how work is produced, I’m interested in how people establish what that work demonstrates.
Our first product, Eleza, helps instructors explore how students explain their own submissions through spoken or typed oral defenses. It connects answers to passages in the work and produces an inspectable evidence dossier. Instructors retain the judgment.
Try Eleza · Explore the source
Grand Prize · Team USA × Google Cloud Hackathon
A multimodal fan experience that matches users to eight athlete archetypes drawn from Olympic and Paralympic history. Built with Gemini, Cloud Run, BigQuery, and Firestore, with dedicated Paralympic archetypes and a visible output-validation trace.
Best of Creative Storytellers · Google Gemini Live Agent Challenge
An AI griot that turns fragments of family history into heritage narratives combining voice, watercolor imagery, ambient sound, and live conversation. Each segment distinguishes historical context, cultural context, and reconstruction.
| Project | What it does |
|---|---|
| Dhamana | A ticketing transaction engine that enforces inventory limits, verified-fan purchase caps, atomic resale, and escrow accounting on Aurora DSQL. Includes a concurrency demo. Demo |
| Hodi | Administers creative consent: scoped permissions, licensing requests, revocation, and auditable decisions across a fleet of agents. Demo |
| Parley | Negotiates for sellers while a deterministic Convex transaction enforces the margin floor. Supports agent-to-agent negotiation over MCP. Demo |
| Karani | Builds rubric-linked evidence sheets from student submissions, validates citations, and escalates uncertainty for instructor review. Includes an offline replay of recorded model runs. |
| Sema | Helps film editors create audio description through visual evidence, measured narration timing, human approval, and export of approved narration and scripts. |
| Akili | Grounds answers from technical PDFs in page coordinates and source text, with explicit confidence tiers and refusal when evidence is insufficient. |
| Dira | Models commitments as a dependency graph, propagates schedule changes, and replans when an attempted repair fails. Interactive evidence |
| Fly Applicant | Simulates a fruit-fly connectome to drive a browser-based job-application scene, with neural activity visualized live. Watch it |
More projects · agent infrastructure & verification
| Project | What it does |
|---|---|
| Amini | Captures agent decision traces, evaluates policies, and organizes violations, incidents, and audit reports. |
| Baraza | Gives agent memory an inspectable record: quoted claims, contradiction review, human ratification, and retractable guidance. |
| Muhuri | Records ordered agent actions, seals them with cryptographic commitments, and anchors proof bundles in S3 Object Lock for offline verification. |
| Thabiti | Meters usage from event logs with deterministic aggregation, sealed billing windows, and independently verifiable receipts. Demo |
| Halisi | Enforces one claim per attested credential using passkey verification, single-use challenges, and atomic DynamoDB writes. |
| Sift | A browser extension that checks whether cited sources support an AI answer’s claims and explains when verification is unavailable. |
| AutoSRE | Diagnoses cloud incidents, supports AWS Lambda alias rollback, and checks recovery before generating an incident report. |
More projects · education, community & adaptable software
| Project | What it does |
|---|---|
| Kijiji Scholars | Supports college planning with academic profiles, roadmaps, scholarship discovery, and versioned essays with AI feedback. |
| Daraja | Builds translation models for underserved language pairs, with a Somali–Swahili training pipeline and an app that runs inference through local Ollama. |
| Zamu | Fills volunteer roster gaps using qualification and workload rankings, explicit authority to act, and verification of the resulting assignment. |
| Asili Agents | An earlier commerce build: a team of agents for micro-sellers, grounded in a live catalog with deterministic pricing and human approval. Demo |
| Sikia | A customizable money dashboard whose runtime agent edits the interface, while a kernel checks changes and commits or rolls them back. |
- IBM · Product Manager Intern · Summer 2026 — Docling and the watsonx platform.
- Oracle Cloud Infrastructure · Software Engineer Intern · Summer 2025 — Built dashboarding infrastructure that closed a 22% coverage gap across 700K+ records.
- BU NSBE — Secured $100K+ in corporate funding as VP of External Affairs and led the chapter to Nationals in Baltimore.
- MARKITIT · Co-Founder & Product Lead — Helping small businesses discover and prepare for pop-up market opportunities.
- MLT Career Prep Fellow · Alpha Phi Alpha Fraternity, Inc., Sigma Chapter
Languages: Python · TypeScript · C++ · Java · SQL
AI & applications: Gemini · Google ADK · Amazon Nova · React · Next.js · FastAPI
Infrastructure & data: Google Cloud · AWS · Oracle Cloud · Convex · PostgreSQL · Firestore · Docker