C++ · Python · AI Systems
M.Tech in Computational & Data Science from NIT Karnataka, with one year of software-development experience at Siemens Digital Industries Software.
Interested in C++, Python, AI systems, automated validation, GPU/ML systems, and performance engineering. Currently seeking entry-level Software Engineering, AI Infrastructure, ML Systems, and C++/Python roles.
A validation-first AI agent for safely modernizing legacy C++, using NVIDIA Nemotron through Nebius Token Factory.
- 20/20 validated patches across TinyXML-2 and pugixml; 95% first-pass success, one strong-model escalation, and 72 offline tests
- Benchmark scope: one pattern, 10 candidates, and one run per repository
- Uses
clang-tidysemantic discovery, bounded, location-anchored model context, and exact-match, location-aware patching - Validates changes with CMake builds, CTest discovery, and tests; rejects zero-test repositories
- Supports fast-to-strong model escalation and rollback of unsuccessful changes
- Records attempt-level latency, failure, revision, CMake, and token telemetry
An experimental LLM inference engine focused on inference internals and GPU performance.
- Achieved approximately 24.9 generated tokens/second on a 4 GB NVIDIA GTX 1050
- Implemented explicit prefill and iterative token decoding with KV caching; served Qwen2.5-0.5B-Instruct through FastAPI
- Measured prefill/decode latency, throughput, token counts, and device-level metrics
Developed during my M.Tech project and Siemens internship to validate visual fidelity between Siemens NX views and exported Technical Data Packages.
- Rendered STL reference views and reconstructed matching viewpoints from 3D PDF camera-to-world matrices
- Compared silhouettes using SSIM, IoU, dilation, and difference maps
- Automated pass/fail checks for visual regression detection
Evaluated technical indicators for next-day stock direction.
- Engineered momentum, volume, volatility, and trend features; compared linear and tree-based classifiers
- Used chronological train/test splits to reduce time-series leakage
- Observed weak AUC performance; results did not establish a reliable predictive signal
June 2025 – June 2026
- Contributed C++ and Python code to the NX Model Based Definition team
- Developed an automated image-comparison pipeline for Technical Data Package validation
- Integrated third-party APIs and implemented backend logic for engineering annotation workflows
- Investigated export, rendering, viewport, and digital-signature defects; built utilities and automated regression tests
- Worked with code reviews and debugging in a large engineering software codebase
- Languages: C++, Python
- AI/ML: PyTorch, Transformers, scikit-learn, LLM inference, KV caching
- Development & validation: Linux, Git, CMake, CTest, clang-tidy, FastAPI, OpenCV
- M.Tech, Computational & Data Science — NIT Karnataka, 2024–2026
- B.Tech, Instrumentation and Control Engineering — VIT Pune, 2019–2023
- GATE CSE 2024: Score 621, AIR 2067