Pranav Unnikrishnan

ML Engineer • Data • CV • Robotics
Experience

Work that moved the needle.

Selected roles across ML, GenAI, Computer Vision, and Applied Research.

Jump to Projects See Publications

Experience

Impact-focused highlights. (You can link each role to a project case study later.)

Updated

Machine Learning Consultant • GTI Energy

Jan 2026 — May 2026 Fort Collins, CO (USA)
  • Processed a 24-month dataset of hourly AMI records for 5,227 customers, integrating weather covariates and transforming 1D temporal data into 2D variations using FFT and 2D convolutional kernels.
  • Developed 5 ML architecture pipelines (TimesNet, XGBoost, DHR, Random Forest, OCSVM) for multi-day anomaly detection, achieving R² up to 0.97.
  • Utilized baseload-excess decomposition and pooled estimation across 275,000+ observations to isolate seasonal heating demand.
  • Engineered an automated inference pipeline to identify consumption irregularities via reconstruction error analysis.

IT Analyst • Office of Information Technology, University of Colorado

Oct 2024 — May 2026 Boulder, CO (USA)
  • Developed a pilot RAG system using LangChain and vector databases, reducing Tier-1 manual ticket volume by ~25%.
  • Performed EDA on historical support ticket data using Python/Pandas to identify network bottlenecks.
  • Architected an automated ticket classification model using LLMs to prioritize issues, achieving an 85% satisfaction rate.
  • Managed enterprise infrastructure and automated routine system audits across campus labs using Bash scripts.

Machine Learning Engineer • Hyperworks Imaging

Mar 2024 — Jul 2024 Bengaluru, KA (India)
  • Engineered an automated pipeline using Detectron2 (PyTorch) to digitize raster well log images with 80% precision.
  • Developed a hybrid Generative + Transformer architecture: GANs for gridline removal and Transformers for precise curve mapping.
  • Architected an OCR-driven extraction system with unsupervised clustering + heuristic post-processing, benchmarking outputs against LAS files via Pandas/NumPy.

Machine Learning Engineer • AI India Innovation Center

Oct 2023 — Jan 2024 Pune, MH (India)
  • Fine-tuned open-source LLMs (PyTorch + Hugging Face) for mathematical problem-solving; improved memory efficiency using 4-bit/8-bit quantization to reduce VRAM footprint.
  • Built a synthetic data generation pipeline via RAG to bridge data gaps, achieving 70% model accuracy.
  • Developed a high-precision OCR system for automated passport extraction; deployed end-to-end solutions using Docker and scalable inference patterns for low-latency serving.

Research Assistant • Amrita School of Engineering

Sep 2022 — Oct 2023 Coimbatore, TN (India)
  • Built and deployed a real-time face mask detection system on NVIDIA Jetson Nano edge devices using transfer learning and TensorRT optimization.
  • Formulated a speech-to-gesture mapping framework using GRU, BiGRU, and multi-head attention to translate acoustic features into 3D human motion representations.
  • Ran time series forecasting experiments for stock market data, benchmarking ARIMA/SARIMAX vs RNN/LSTM/Transformer models.
  • Extracted and preprocessed 122 years of historical daily gridded rainfall data from the India Meteorological Department (IMD).
  • Co-authored three peer-reviewed conference papers in Springer book series (LNNS and CCIS), presenting results at international computing conferences.