Hey everyone, my name is Sid and I'm a Junior in the UC Berkeley M.E.T. program — dual degree in EECS and Business Administration (Haas). I work at the intersection of on-device ML, agentic systems, and developer tooling, with a focus on making capable models run efficiently on consumer hardware.

Experience

RunAnywhere (YC W26) — Machine Learning Engineer

• Building an on-device speech-to-tool-call pipeline that runs entirely on consumer hardware. Whisper for ASR, a quantized Qwen3-0.6B (NF4 + LoRA) for intent parsing, and a hot-swappable LoRA adapter system implemented in C++ that enables sub-second domain switching without reloading the base model. Revyl (YC W24) — Software Engineering Intern
• Built a Model Context Protocol (MCP) server that orchestrates mobile end-to-end test runs, reducing p95 latency by 40% through Redis-backed result caching and request batching. Designed and deployed a Slackbot that translates natural-language queries into SQL against the observability backend, replacing the team's Grafana triage workflow and cutting incident-response time. URAP (Berkeley Haas) — Undergraduate Researcher
• Research under Prof. Anastassia Fedyk on AI propaganda detection, fine-tuning LLaMA 3.1 on GDELT news data to identify coordinated narrative campaigns across global news sources.

Projects

• Remixify (Won Best Use of Gemini, CalHacks): AI-powered music remixing platform. Won the Gemini track at CalHacks against a competitive field.
• Stryd: Fitness watch with on-device exercise detection. ARM Cortex-M4F driving a 6-axis IMU over DMA SPI, complementary-filter sensor fusion, and a quantized 1D-CNN running locally for activity classification. Streams telemetry to an iOS companion app over BLE 5.0 GATT.
• NutriGenie: Built an AI nutrition platform that aligns macros, budgets, and fitness goals against live grocery store inventory and pricing. Scaled to 300+ active users.

Leadership

• Senior Mentor, CS Mentors Berkeley — Teach weekly sections for CS 61B (data structures) and CS 61C (computer architecture: RISC-V, caches, virtual memory, concurrency).
• Founder, Shape-Up-With-Sid — Youth fitness nonprofit serving underserved communities.

Technical Skills

• Languages: Python, C++, TypeScript, Swift, RISC-V Assembly
• ML and Systems: PyTorch, llama.cpp, ONNX, LoRA / QLoRA, Whisper, NVIDIA NeMo
• Backend and Web: FastAPI, Next.js, React, PostgreSQL, Redis
• Infrastructure: Docker, AWS (Lambda, EC2), Vercel, Anthropic SDK, OpenAI API

Contact

• LinkedIn: linkedin.com/in/siddharth-veerapaneni
• Email: sidv@berkeley.edu