root@basav:~# init_sequence --verbose

AI_ARCHITECT :: READY

> SPECIALIZATION: [COMPUTER_VISION, GENERATIVE_AI, REAL-TIME_INFERENCE]
> FOCUS: production ML systems — efficient models, robust pipelines, low-latency deployment
> STATUS: open to collaborations and engineering roles

I'm an AI architect who builds reliable, production-ready computer vision and generative systems. I design end-to-end solutions—data collection, model development, and optimized deployment—with a focus on latency, robustness, and clear real-world impact. AWS Certified ML Associate; curious about research-driven approaches and practical experimentation.

DOWNLOAD_CV.exe VIEW_PROJECTS.sh CONTACT_ME
Skills:
PyTorch
Computer Vision
Backend (Python)
AWS / Deployment
Optimization
Generative Models

/// EXECUTION_LOG (EXPERIENCE)

[2025.04 - PRESENT] VALERE_LABS

SOFTWARE_DEV :: AI_ML

  • > Leading development of AI systems for sports analytics, integrating deep learning with biomechanical simulation workflows.
  • > Engineered a multi-stage 3D human-motion reconstruction pipeline that lifted 2D keypoints to 3D poses, fused temporal cues using SLAM-style optimization, and generated SMPL-X meshes for downstream analysis in OpenSim.
  • > Designed and optimized the video-processing pipeline, reducing end-to-end latency by 55% through parallelization, batching, and caching.
  • > Fine-tuned and deployed YOLO-based detection and tracking models on edge/mobile; achieved major speedups using ONNX Runtime, TensorRT, and custom pre/post-processing kernels.
  • > Performed model optimization via quantization, pruning, and operator-level optimizations, improving inference throughput by nearly 50% across devices.
  • > Contributed to internal GenAI systems, integrating LLMs, RAG, and agent-based automation for multimodal query workflows.
[2024.05 - 2025.03] MONOTYPE_IMAGING

AI_TRAINEE :: RESEARCH

  • > Developed and trained a custom diffusion model for in-image text editing, enabling localized text modification while preserving font characteristics, layout structure, and visual coherence.
  • > Conducted applied research on Rectified Flow and ControlNet to enhance controllability, structural consistency, and text fidelity in generated outputs.
  • > Fine-tuned large-scale models (SDXL, Flux) on proprietary font datasets, significantly improving stylistic consistency.
  • > Created high-quality synthetic datasets and fine-tuned segmentation models to support font-conditioned diffusion pipelines.
  • > Accelerated inference by 65% through optimization of generation stages, scheduler parameters, and post-processing.
[2023.07 - 2024.05] HUMANLI.AI

ML_ENGINEER

  • > Built a custom CNN for pneumonia detection from chest X-rays, achieving 95% accuracy and advancing to clinical trial validation.
  • > Designed and deployed an LLM-based analytics workflow using LLaMA-2, RAG, and multi-agent orchestration.
  • > Led dataset preparation, cleaning, augmentation, and model validation for medical imaging pipelines.

/// INSTALLED_PACKAGES (SKILLS)

LANGUAGES

PYTHON C++ JAVA C

AI_CORE

PYTORCH TENSORFLOW OPENCV SCIKIT-LEARN HUGGING_FACE KERAS LANGCHAIN

SYSTEMS

DOCKER GCP LINUX GIT IOT (R-PI/ARDUINO)

CERTIFICATIONS

AWS ML Badge AWS Certified Machine Learning - Associate [VIEW_CREDENTIAL]

/// PROJECT_DATABASE

IMAGE_FORGERY_DETECTION — Case Study

Problem: Identify manipulated images reliably in the wild. Approach: Trained a CNN ensemble with patch-level augmentation and post-processing heuristics. Role: End-to-end prototype from dataset curation to evaluation.

IMAGE_CAPTION_GEN — Case Study

Problem: Create descriptive captions for images suitable for accessibility. Approach: Hybrid CNN encoder + LSTM decoder, pretrained embeddings and fine-tuning. Role: Model design and deployment prototype.

Image Forgery
Analyzing...

IMAGE_FORGERY_DETECTION

Detecting image forgery using CNN (Keras/TensorFlow). Streamlit frontend.

ACCESS_REPO >>
Image Caption
Analyzing...

IMAGE_CAPTION_GEN

Hybrid Model (CNN + LSTM) for generating captions. Django backend.

ACCESS_REPO >>
Sign Language
Analyzing...

SIGN_LANG_DETECTION

Real-time detection using CNN & OpenCV.

ACCESS_REPO >>
Recommender
Analyzing...

MOVIE_RECOMMENDER

Collaborative + Content-based filtering hybrid system.

ACCESS_REPO >>
Churn Pred
Analyzing...

CHURN_PREDICTION

Customer churn analytics using Logistic Regression & Decision Trees.

ACCESS_REPO >>
VIEW_FULL_ARCHIVE >>

/// RESEARCH_ARCHIVE (PUBLICATIONS)

PAPER_01 MARCH 2023

Image Forgery Detection: A Survey Of Recent Approaches

Proceedings of 4 National Research Scholars MEET-2023

Author: Amneet Kaur, Basav Bamrah, Prof. Jaspreet Kaur.

PAPER_02 2022

Sign Language Recognition Using Convolutional Neural Networks

Journal of Advanced Research in Electronics Engineering and Technology (Vol 9 No 1&2)

Author: Basav Bamrah, Prof. Jaspreet Kaur.

/// ESTABLISH_UPLINK

User is currently open to new opportunities. Transmission lines are open.

> SEND_EMAIL_PACKET > CONNECT_LINKEDIN > TWITTER_FEED

Or manually copy address:

bamrah.basav@gmail.com