Rohan Saha
Software Engineer
Hey!
Computer Science @ University of Waterloo | Founder | Ex-varsity athlete (soccer & track) | AI Engineer
Currently learning computer vision, training Muay Thai (kickboxing), and building a course scheduler for UWaterloo.
Co-Founder & CEO
Optimized API performance through PostgreSQL indexing and Redis caching, reducing p95 latency from 220ms to 95ms (57% improvement) and achieving 3x read-query headroom for future scale.
Show More- Drove user acquisition to 2,000+ page views through targeted outreach reaching 15,000+ students, achieving a 12% engagement rate and establishing product-market fit.
GenAI Intern
Built an AWS-deployed multi-agent application with RAG toolchain integration for ML solution execution and root-cause analysis, reducing end-to-end workflow time by 50% and saving 20+ hours per week.
Show More- Designed and implemented a RAG system automating document chunking, preprocessing, embedding generation, and hybrid vector/graph search, reducing data retrieval time by 70% for unstructured data.
- Delivered technical demos to Microsoft, Amazon and Ab Initio engineering teams for enterprise adoption.
Machine Learning Engineer
Built and deployed an AI tutoring platform serving 200+ students in MTE 182, architecting 5 microservices handling chat, RAG, PDF processing, and multi-agent orchestration in Docker.
Show More- Designed a RAG pipeline with custom chunking (800-char, 18.75% overlap), pgvector HNSW indexing for sub-second search, and streaming responses, reducing query latency by 35%.
- Built a multi-agent system using LangGraph with typed state management and dynamic routing, leveraging the Gemini API, and developed a standalone ReAct microservice for mathematical computation.
Backend Team Lead
Led a cross-functional team of 10 engineers and designers to build Electriumap, a crowdsourced platform enabling micromobility riders to locate and share verified charging stations across 200+ urban routes.
Show More- Delivered a real-time, responsive map interface using NextJS, Tailwind CSS, shadcn/ui, and the Mapbox API, achieving sub-300ms load latency and improving geolocation accuracy by 40% through tile caching.
- Optimized routing performance by implementing a corridor-based greedy algorithm integrated with Firebase and Firestore, reducing average detour distance by 96% across 500+ simulated route trials.
PC Optimization
Boosted system performance for 50+ clients using advanced hardware and software tuning.
Show More- Increased frame rates by 60%, reduced CPU/GPU temps by 20%, and cut latency by 50%.
- Used techniques like registry edits, BIOS tuning, GPU overclocking, and BCDEdit tweaks.
Languages
Frameworks & Libraries
Developer Tools
Machine Learning & Data
Awards & Honors
Best Cruptocurrency Hack
HackCanada 2025
President's scholarship of Distinction
University of Waterloo
IntelligentForest
Forest Fire Prediction Model
DigiTag
Blockchain based E-commerce platform.
VisionTeX
Image to Text converter for LaTeX
Swipe Chaser
Modular 2D Engine Game.