Hi There,
I'm Aloukik Aditya

i am

About Me

About Me

I'm Aloukik

Head of AI | GenAI Expert | AI Agents

AI Engineer & Architect with 8+ years building $MM impact solutions. I architect AI systems that replace entire departments and unlock new revenue streams. Currently leading breakthrough AI initiatives that major tech labs said were impossible.

From model building to production systems handling billions of requests. From training custom LLMs on distributed GPU clusters to deploying edge AI on hospital robots, I own the complete AI lifecycle. Specializing in multimodal AI systems with Flux, Whisper, custom TTS, and cross-modal transformations.

Name: Aloukik Aditya

Phone: +1 905 906 0188

Email: aloukikaditya@gmail.com

Location: Toronto, Ontario, Canada

Resume

Skills & Expertise

GPT-4/OpenAI
Llama 3.1
Gemini
Claude 3
Stable Diffusion XL
Midjourney/DALL-E
Python
PyTorch
TensorFlow
AWS SageMaker
GCP Vertex AI
Azure ML
Docker
Kubernetes
LangChain
NumPy
Pandas
Scikit-learn
OpenCV
YOLO/Detectron2
Whisper/TTS
Pinecone/ChromaDB
MLflow/MLOps
CUDA/TensorRT

8+ years of Experience

Stealth Startup | May 2023 - Present

Head of Artificial Intelligence

  • Built AI achieving what $50M labs couldn't. 10K+ users daily, 95% cheaper than GPT-4, patent-pending innovations.
  • Multimodal AI Systems: Building agents with Flux, Whisper, custom TTS, image/video generation • End-to-end fine-tuning pipelines • Cross-modal transformations (text↔image↔video)
  • Production deployment of Llama 3.1, GPT-4, Claude 3, Mistral, Gemini. RAG with Pinecone, ChromaDB, Weaviate, Qdrant.
  • Fine-tuning with LoRA, QLoRA, PEFT, RLHF. Custom training pipelines for voice cloning and hyper-realistic face generation.
  • Infrastructure: AWS (SageMaker, Bedrock), GCP (Vertex AI, TPUs), Docker, Kubernetes, CUDA optimization, Edge AI deployment.

Tiger Analytics | Jan 2022 - May 2023

Principal Machine Learning Engineer & Strategic Consultant

  • PepsiCo: Optimized $2B marketing portfolio. Built ML system achieving 300% improvement in marketing efficiency, $20M saved.
  • Analyzed 10M+ reviews using BERT, RoBERTa, T5. Connected sentiment patterns to purchase behavior leading to $50M product reformulation.
  • Deployed LLMs before ChatGPT for Zebra Technologies' contract intelligence. Review time reduced from 6hr to 3min.
  • Document AI Pipeline: OCR + LayoutLM for 50K PDFs monthly. Built LangChain RAG with Pinecone. $5M annual savings achieved.
  • Production Scale: AWS SageMaker, Docker, Kubernetes, MLflow. Real-time dashboards used in weekly CMO meetings.

Solaris Robots | June 2021 - Jan 2022

Lead Computer Vision Architect

  • Deployed infection-fighting robots in hospitals with AI vision for 10x less cost than $50K fixed systems.
  • Edge AI: Squeezed YOLOv4, SSD MobileNet onto NVIDIA Jetson Nano/Xavier. Achieved 45 FPS real-time detection.
  • 18x faster pipeline through PyTorch → ONNX → TensorRT optimization. Models 10x smaller, equally accurate.
  • Built ROS nodes for autonomous navigation + computer vision. Robots detected PPE compliance, monitored sanitization.
  • Real Impact: 5 hospitals, 73% better risk detection, 4 hours saved daily per staff.

Lakehead University | Aug 2020 - Jun 2021

Senior Principal Research Scientist

  • Built AI that predicts accidents 3.2 seconds before they happen using single camera setup.
  • Generated 455,240 dangerous driving scenarios using GTA 5 instead of $10M crash datasets. Published approach now industry standard.
  • Model ensemble (SENet, ResNeXt, EfficientNet) achieving 0.915 ROC-AUC. Custom attention mechanisms + transfer learning.
  • ComputeCanada clusters, 128 GPUs, distributed PyTorch training. 500GB dataset processed in 1 week vs 6 months locally.
  • IEEE SMC 2021 Best Paper nominee. Tesla engineers cited our paper. Open-sourced everything.

Google | Jun 2020 - Aug 2020

Research Scientist

  • Made Google 20% better at content safety. In production for 8.5B daily searches.
  • TensorFlow/PyTorch hybrid CNN achieving 94% accuracy (up from 85%). 20% faster via INT8 quantization.
  • Model compression: 250MB → 50MB through pruning. Added OCR pipeline for text detection.
  • Deployed to 1% traffic (85M queries/day) then full scale. Prevented ~10M inappropriate exposures quarterly.
  • 100ms → 80ms inference optimization. Saved Google $1M+ annually. Still running in production.

Fanshala | May 2018 - Sept 2019

Principal AI Architect & Technical Director

  • Built AI that knows what users want before they do. CTR improved from 0.8% to 2.4%.
  • LDA topic modeling processing 1M+ pages/day. Custom TF-IDF vectorizer weighing 10K+ keywords.
  • SVM + Linear Regression predicting user bounce with 89% accuracy. Real-time processing pipeline.
  • Containerized ML before DevOps was cool. Scaled 100x without hiring. Revenue increased 250%.
  • 50+ category classification. Bounce rate reduced from 70% to 42%. Users actually happy.

IIM Lucknow | Jan 2018 - Feb 2018

Data Scientist Intern

  • Mastered R Language, Data Analytics, and Econometrics for business insights and anomaly detection.
  • Handled missing values using KNN on car dataset, achieving 0.12 MSE in predicting fuel type.
  • Achieved 91.02% classification accuracy using cosine and Manhattan distances for data point comparisons.

Key Projects & Impact

Enterprise AI Platform (Stealth)

Built AI achieving what $50M labs couldn't. 10K+ daily users, 95% cheaper than GPT-4. Patent-pending multimodal AI with custom LLMs, voice cloning, and image generation.

Confidential

PepsiCo Marketing Optimization

Optimized $2B marketing portfolio across 50+ brands. Achieved 300% improvement in marketing efficiency, saving $20M. Analyzed 10M+ reviews leading to $50M product reformulation.

Enterprise

Collision Detection AI (Published)

IEEE SMC 2021 published research. AI predicts accidents 3.2 seconds before they happen. Used GTA 5 for synthetic data generation. Tesla engineers cited our work.

View Paper Code

Hospital Safety Robots

Deployed infection-fighting robots in 5 hospitals. 73% better risk detection using edge AI on NVIDIA Jetson. Real-time PPE compliance monitoring saving 4 hours daily per staff.

Healthcare AI

Google Search Safety

Improved content safety by 20% for 8.5B daily searches. Model optimization saved $1M+ annually. Prevented ~10M inappropriate exposures quarterly. Still in production.

Production Impact

AdTech Mind Reader

Built AI predicting user behavior with 89% accuracy. CTR improved 3x (0.8% → 2.4%), bounce rate down 40%, revenue up 250%. Scaled 100x without additional hiring.

Revenue Driver

Education

Education is not the learning of facts, but the training of the mind to think.

Master's degree, Computer Science

Lakehead University | Thunder Bay, Canada

Graduated | Research Focus: Computer Vision & AI Safety

Bachelor's degree, Computer Science

Guru Gobind Singh Indraprastha University

Graduated | Focus: Machine Learning & Data Science

Get in Touch

Aloukik's Portfolio

AI Engineer & Architect with 8+ years building enterprise-scale ML systems. From MVP to production at Fortune 500 scale. Let's build something transformative together! 🚀

quick links

home skills experience projects education about

contact info

+1 905 906 0188

aloukikaditya@gmail.com

Toronto, Ontario, Canada

Designed with by Aloukik Aditya