Experience

• Client-facing post-deployment work: partner directly with enterprise customers to develop custom solutions on top of the deployed platform, then migrate those solutions onto client systems for a clean handoff into their environment.
• Architected and deployed 3 scalable AI microservices on IBM Granite 3.3 and Llama 3.3, implementing RAG patterns that process 1M+ queries at 87% answer relevance for enterprise IT and HR support agents.
• Optimized LLM inference with Redis-based prompt caching and request batching — 40% less API overhead while holding <1.5s P95 latency.
• Built automated evaluation pipelines benchmarking accuracy, latency, and cost-per-token across model variants, cutting model selection cycle time 35% and directly informing deployment decisions for Fortune 500 clients.

• 1st Place Penn State Research Symposium Poster Competition (April 2026) for electrochemical biosensor automation platform research.
• Architected a multi-MCU embedded system decoupling robot/pipette control, ADC current sampling, and DAC waveform generation across three dedicated STM32s — resolving RTOS scheduling conflicts that prevented reliable simultaneous operation and enabling precise high-frequency LAACV acquisition.
• Engineered a closed-loop volume calibration system in Python correcting nonlinear peristaltic pump response and creep drift, recovering dispensing accuracy across extended multi-day runs; CV + LAACV multimodal fusion improved limit of detection 100×.
• Implemented an LSTM predictive model (TensorFlow) forecasting sensor degradation at 89% accuracy, enabling proactive maintenance and extending average sensor lifetime by 2 weeks.

• Led an 8-person executive board for a 150+ member chapter, growing active project participation 150% and launching embedded systems workshops reaching 50+ students.
• Secured $5,000 in corporate sponsorship to launch a university-wide robotics competition attracting 10 teams across 5 departments.