hi, I'm
Education: Williams College
Computer Science, Number Theory, Creative Writing, Filmmaking
Studio 275
Conversations about student career journeys, tech, and building in public
BrijHowardSarin
Strategy & Product Intern, Jane Street
MiltonVento
Amazon SDE Intern
TashriqueAhmed
Meta SWE Intern
Projects
Carpooling for Remote Colleges
Tech Stack
React NativeExpoNext.jsTypeScriptPostgreSQLStripeClerk
Key Features
- •Real-time ride matching and booking
- •Secure payment processing with Stripe Connect
- •Chat messaging between drivers and riders
- •Location tracking and route optimization
Training multilingual language models for underrepresented African languages
Tech Stack
PythonPyTorchTransformersDVCW&BSentencePiece
Key Features
- •Supports 6 African languages (Amharic, Fulani, Hausa, Somali, Swahili, Yoruba)
- •GPT-2 architecture with 768-dimensional embeddings
- •16K+ shared vocabulary using BPE tokenization
- •Comprehensive training pipeline with W&B monitoring
Sub-200ms fraud detection pipeline using stream processing and pattern matching
Tech Stack
Java 11Apache FlinkKafkaDynamoDBDockerAWS Lambda
Key Features
- •Complex Event Processing (CEP) for fraud pattern detection
- •Handles 1M+ events/sec with <200ms P99 latency
- •Stateful stream processing with RocksDB checkpoints
- •Serverless alerting via Lambda and SNS
Research
Explainable AI for Chess Engines
Ongoing
- • Developing natural language explanation systems that translate chess engine evaluations into human-understandable strategic concepts, bridging the gap between computational analysis and human comprehension
- • Implementing attention-based neural architectures to identify and visualize critical board features that influence positional evaluations, enabling transparent decision-making in chess AI
- • Creating interpretable feature extraction pipelines that decompose engine evaluations into fundamental chess principles (material, mobility, king safety, pawn structure) with quantifiable importance scores
Multilingual African Language Model (MALM)
Ongoing • Williamstown, MA
- • Developing a comprehensive language model supporting 6 major African languages (Amharic, Fulani, Hausa, Somali, Swahili, and Yoruba) with a unified 16K token vocabulary
- • Engineering robust data pipelines that process and clean multilingual corpora from OSCAR and CC-100 datasets, implementing advanced filtering mechanisms to ensure high-quality training data
- • Optimizing model training across diverse hardware architectures, from NVIDIA GPUs to Apple Silicon, to democratize access to AI technologies for underrepresented African language communities
Topological Graph Theory
Ongoing
- • Investigating graph embeddings on non-planar surfaces, with focus on minimizing crossing numbers and understanding structural properties of graph minors
- • Exploring algorithmic applications of Robertson-Seymour theory and its implications for computational complexity in graph algorithms
- • Connecting spatial graph theory with knot theory to develop new geometric insights and visualization techniques for complex graph structures