Skip to content
View Ishaan-Ansari's full-sized avatar
🎯
Focusing
🎯
Focusing
4 followers · 14 following

Achievements

Achievement: Pull SharkAchievement: Quickdraw

Achievements

Achievement: Pull SharkAchievement: Quickdraw

Block or report Ishaan-Ansari

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Ishaan-Ansari/README.md

Ishaan Ansari

Machine Learning Engineer · Search, Retrieval, Recommendation Systems & GenAI Solutions

I help enterprises architect and deploy production-grade AI systems at scale.

First Principles

def architect_ai_system(business_requirements):
    """
    Lessons from shipping AI to production:
    1. Start with the data pipeline (garbage in, garbage out)
    2. Vector databases are the new relational databases
    3. Prompt engineering is systematic, not magic
    4. Always have a fallback to traditional algorithms
    5. Monitor for drift, hallucinations, and edge cases
    """
    if requires_compliance():
        return multi_layer_validation() + audit_trails()
    elif requires_scale():
        return async_processing() + caching_layers()
    return simple_and_maintainable()

Building

Project Description
MIRAGE Multimodal RAG framework that intelligently integrates visual embeddings from medical images with retrieved clinical knowledge, leveraging dynamic prompt control to enhance factual precision and interpretability in medical reasoning tasks.
Claimova Automation engine leveraging document AI and decision workflows to reduce denial rates and accelerate payout pipelines
DeepSeek V3 100M+ param transformer with Multi-Head Latent Attention, MoE, and Multi-Token Prediction, trained on 2.5B tokens of FineWeb-Edu.
Aegis Multimodal Video Compliance QA Pipeline to audit content against regulatory standards. Following LLMOps best practices
Nudge "Nudge" is a recommendation engine to provide personalized product recommendations based on user behavior and preferences.
Style Transfer Deep learning optimization pipeline that leverages convolutional feature representations to synthesize artistic styles onto arbitrary images.
Captionix Generating Captions for images using CNN & LSTM and attention.
CardioVision Project aims to detect cardiovascular diseases by classifying Electrocardiogram (ECG) images.

Field Notes

Repo What it is about?
Machine Learning The purpose of this project is to provide and learn a deeper understanding of ML fundamentals.
Activation functions It's an ever-evolving project, Implemented more than 10 activation functions and their derivatives. With more updates planned.
Deep Learning This project aims to provide a deeper understanding of how neural networks function internally, without relying on high-level libraries.
Designing Data Intensive Applications My reading notes following "Designing Data Intensive Applications" (DDIA) by Martin Kleppmann
AI Powered Search My reading notes following " AI-Powered Search" by Doug Turnbull, Max Irwin, and Trey Grainger.
Tokenizer This project demonstrates word-level, character-level & Byte-Pair encoding tokenization, assigning token IDs and visualizing them with custom colors.
LLM from Scratch This repository provides a step-by-step implementation of a Large Language Model (LLM) from scratch, covering data preparation, model building, pretraining, and fine-tuning.
Inference Engineering A hands-on series on making LLM inference faster. It covers the theory, the systems reality, and runnable code you can reproduce on a single rented GPU for about a dollar.
MLOps Primer Practical guide to build end-to-end machine learning pipeline and deploy your model in production.
AWS Resource Cleanup Guide While working with AWS, ensure you keep track of all the running and discarded services.

Writing · Gradient Ascent 101

Signal over noise. Deep dives into production AI and systems architecture:

Background
Now ML Engineer II at BOLD (Monster.com & CareerBuilder.com): Building personalized search and recommendation systems. Integrating GenAI (LLMs) to optimize job matching and connect candidates with relevant career opportunities.
Before As an ML Engineer, I worked on building yHealth, an LLM-powered hyper-personalized meals, medicine, and overall lifestyle application, specifically for diabetic patients · Engineered and deployed 4 core features from scratch.

Stack: Python · PyTorch · Vector Search / OpenSearch · LangGraph & LangChain · AWS (Lambda, SQS, Fargate, RDS) · Kubernetes

Connect With Me

Pinned Loading

  1. Data-Structures-and-Algorithms Data-Structures-and-Algorithms Public

    Dedicated repo for all algorithmic problems I tackle, serving as a comprehensive resource for my journey in mastering data structures and algorithms

    Python

  2. Deep-Learning-from-scratch Deep-Learning-from-scratch Public

    Implementing deep learning architectures from scratch

    Jupyter Notebook

  3. Finetuning-LLMs Finetuning-LLMs Public

    Workflows for fine-tuning various Large Language Models (LLMs) on task-specific datasets

    Jupyter Notebook

  4. LLM-from-scratch LLM-from-scratch Public

    Implemented LLM from Scratch

  5. DeepSeek-from-Scratch DeepSeek-from-Scratch Public

    This repository contains an ongoing implementation of DeepSeek from scratch. Currently, it focuses on building core components such as Multi-Head Latent Attention (MLA) — a memory-efficient alterna…

  6. Multi-Modal-Compliance-Orchestration-Engine Multi-Modal-Compliance-Orchestration-Engine Public

    Automated Video Compliance QA Pipeline to audit content against regulatory standards. Following LLMOps best practices

    Python