[blog] [github] [linkedin] [resume]

Yashwanth Allakky

Hyderabad, India
Yashwanth Allakky

Hi, I'm Yashwanth.

I'm a machine learning engineer that loves building systems from the ground up.

I love training neural networks that have an impact on real life.

I'm currently training large reinforcement learning agents that save electricity at Zodhya. I also fiddle with LLMs both professionally and in my personal projects.

Prior to this, I studied AI and ML at UoH and computer science at JNTUH.

In my free time I like to play the drums, carve, draw, lift weights and run.

This website serves as a sample of what I am currently building, learning and thinking about.

// experience

Machine Learning Engineer
Zodhya
2021 - Present
  • As a founding ML engineer, I architected and implemented deep learning solutions that expanded product capabilities from small retail spaces to large-scale commercial buildings and airports.
  • Designed and deployed ML systems handling 100+ real-time sensor inputs per minute across multiple HVAC zones, achieving 30% reduction in energy consumption, ~250,000 electricity units annually.
  • Built internal deep learning toolkit using PyTorch and OpenAI Gym that reduced experiment cycle time from weeks to days and standardized training across the team.
  • Built an RAG for automated report generation using LLMs that processes real-time energy monitoring data to generate instant insights and answer ad-hoc client queries, eliminating manual report generation.
Machine Learning Engineer
Consultant
2020 - 2022
  • Developed a novel NLP-based sequence matching system for analyzing urban mobility patterns across 6500+ individuals, achieving 92% accuracy in predicting daily activity patterns.
  • Engineered features incorporating socio-economic factors, spatial data, and temporal patterns to optimize urban mobility predictions for a major Indian city simulation project.

// education

University of Hyderabad
PG Diploma in AI and ML
2021 - 2022
Jawaharlal Nehru Technological University
BTech in Computer Science
2016 - 2020

// projects

Bazaz Dysphagia Prediction
Ongoing
  • Research project with Weill Cornell Medicine for predicting new Onset Dysphagia post-ARS. Built a novel risk-scoring model to predict the presence of dysphagia with 83% sensitivity.
Malware Classification using Machine Learning
2021
  • Built a classifier to detect malicious software using Microsoft’s malware dataset. Engineered novel features from ASM and operation code files including n-gram patterns and instruction frequencies. Processed and analyzed over 120GB of binary data to achieve 0.008 log loss through extensive feature engineering and gradient boosting techniques

// achievements

Winner
Smart India Hackathon, Ministry of Education
2019
All India Rank 81
IEEE Xtreme, IEEE
2019

If what you see interests you, please get in touch. I'm always open to collaborating on projects that interest me.

mail@yash.fr