class Harikeshav:
def __init__(self):
self.role = "Systems & AI Engineer"
self.education = "B.S. Computer Science @ Ohio State (Accelerated 3-Year Track)"
self.grad = "May 2027 · Dean's List & University Honors (all semesters)"
self.focus = ["High-Performance Computing", "Distributed Systems",
"LLM Architecture", "Applied Cryptography"]
self.currently = "SWE Intern @ GE Aerospace — AI / FinFlow Team"
def philosophy(self) -> str:
return "Go a layer deeper — from SIMD kernels to multi-agent orchestration."- 🧪 Researching encrypted computation — compiling neural nets to run directly on ciphertext with Fully Homomorphic Encryption.
- 🏆 7× hackathon-winning teams (RevolutionUC, TartanHacks @ CMU, NextHacks, HackOHI/O).
- 📫 Reach me at r.harikeshav@icloud.com · more at harikeshav.me
Languages
AI / ML & LLMs
Backend & Web
Cloud, DevOps & Data
Systems & Security
-0071C5?style=for-the-badge&logo=intel&logoColor=white){kind=icon}
-4B275F?style=for-the-badge&logo=letsencrypt&logoColor=white){kind=icon}
| When | Where | What |
|---|---|---|
| 2026 | GE Aerospace — AI / FinFlow | Built a Bedrock + Claude extraction pipeline cutting per-doc latency 74%; drove mypy --strict errors 93 → 0 and grew the test suite to 78 tests. |
| 2025 | Siage Solutions | Shipped a RAG knowledge assistant over 100+ manuals (15M+ tokens), cutting search latency 91%; served vLLM inference to 2,500+ daily users. |
| 2023 | IIT Madras — Wireless & Spatial AI | Modeled indoor Wi-Fi signal strength to R² = 0.975, cutting manual site-survey effort 86%. |
- 🥈 2nd Place · Medpace Track · RevolutionUC 2026 (60+ teams) — Pulse, a multi-agent clinical-trial safety platform
- 🥉 3rd Place · Visa Track · TartanHacks 2026 (CMU, 1,200+ hackers) — Penny, an AI personal-finance assistant
- 🥉 3rd Place · The Token Company Track · NextHacks 2026 (2,000+ participants) — Distill, an LLM context-compression engine
- 🎖️ Best AI Hack Runner-Up · HackOHI/O 2025 (800+ participants) — LeadForge, an autonomous agentic sales platform
