class AmanPant:
    name       = "Aman Pant"
    location   = "IIT Mandi, Himachal Pradesh 🇮🇳"
    role       = "AI Software Engineer Intern @ Bosch"
    education  = "M.Tech CSE @ IIT Mandi (CGPA 8.67)", "B.Tech DevOps @ UPES (CGPA 8.35)"
    focus      = "Generative AI", "Diffusion Models", "LLMs", "Agentic Systems", "Document Intelligence"
    gate_2024  = "AIR 1380 | 98.88 Percentile | Score 661 🏆"
    leader     = "Leading ₹1 Cr AmEx research + 100+ placement coordinators at IIT Mandi 😎"

🔴 AI Software Engineer Intern — Bosch Global Software Technologies (Dec 2025 – Present)

• Built end-to-end agentic Document Intelligence pipeline (LangChain + GLM-OCR + QWEN) — achieved 88.98% recall (vs GPT's 85.04%)
• Config-driven orchestration scaling across document types without retraining; benchmarked GPT vs 10+ open-source OCR/LLM models

⚙️ DevOps Engineer Intern — Xebia (Jun 2022 – Aug 2022)

• Containerized Node.js + MongoDB service: 1,200+ daily bookings, P95 latency cut by 40%
• Jenkins CI/CD: release time 2 hrs → 15 min, deploy frequency 4×

📄 Agentic Document Intelligence  ·  Bosch · Dec 2025–Present

• End-to-end pipeline (LangChain + GLM-OCR + QWEN) extracting structured JSON from table-heavy PDFs
• Hit 88.98% recall on shipping-bill benchmarks — on par with GPT (85.04%), at a fraction of the cost

📊 Synthetic Financial Data — American Express  ·  Jun 2025–Present

₹1 Crore research collaboration on fraud & default detection

• Benchmarked CTGAN, TVAE, FinDiff & TabSyn on fidelity (Shape, Trend), utility (C2ST), and privacy (α-Precision, β-Recall, DCR)
• Diffusion model achieved 95.89 C2ST · 96.49 α-Precision, DCR near the 50% privacy threshold

🧠 Brain Tumor Progression  ·  Jan–Jun 2025

• Automated BraTS-Pro pipeline for longitudinal MRI tumour progression analysis
• Siamese 3D-CNN achieving 88% accuracy & 90% recall, built to support reliable clinical assessments over time

🎙️ Voice of the Nation — Language Identification  ·  Apr 2025

🥈 2nd place among 300+ teams — HCL AI Hackathon @ IIT Mandi

• 10-language speech classifier using Wav2Vec XLS-R 300M + OpenL3 with Orthogonal Projection Loss → 82% accuracy
• SVM gender classification module → 99% accuracy

⚡ Hybrid Cache Replacement Policy  ·  Oct–Dec 2024

• Smart block-replacement policy in MacSIM for hybrid volatile/non-volatile cache systems
• 45% reduction in power consumption vs SRAM-only caches

Languages

AI / ML / Deep Learning

Agentic AI / LLMs

RAG Systems

Cloud / DevOps

Databases & Retrieval

• 🎯 GATE 2024 — AIR 1380 | 98.88 percentile | Score 661
• 🥈 2nd Place — Voice of the Nation AI Hackathon (300+ teams, HCL × IIT Mandi)
• 💻 Codeforces — Peak rank 256 | 1000+ problems solved
• 🎓 Graduated top 1% of UPES B.Tech 2023 cohort
• 📋 Overall PG Placement Coordinator @ IIT Mandi — leading 100+ coordinators
• 🧑‍🏫 Teaching Assistant — CS 212: Design & Analysis of Algorithms (300+ students)

"The best time to start was yesterday. The second best time is now."