Instagram story viewer> @priyal.py> Posts

Priyal | DS & ML

@priyal.py

432

posts

140.2K

followers

189

following

🤖Large Language Models | AI | Data Science
👩‍💻Digital Progress Diary
📊Let’s dive into tech together

POSTS STORIES REELS TAGGED

Download All

resources to follow alongside this roadmap:1) Krish Naik – ML, DL, NLP, MLOps, end-to-end projects2) CampusX – strong fundamentals & beginner friendly ML3) freeCodeCamp – structured full-length courses4) MLOps Zoomcamp – industry-grade MLOps (CI/CD, deployment, monitoring)4) LangChain & LlamaIndex playlists These resources are more than enough if you:-follow a roadmap-build projects-deploy what you buildYou don’t need 100 sources.You need consistency + execution.Share it with someone starting their AI journey.#AIEngineer #MachineLearning #DeepLearning #MLOps #GenerativeAI #DataScience by @priyal.py

38k

185

5 months ago

Putting it all together, when you write PyTorch code, PyTorch launches CUDA kernels, the CUDA Toolkit provides the tools and libraries required for execution, the CUDA Driver communicates with the GPU hardware, and the GPU performs the computation in parallel across thousands of cores.#learningtogether #womeninstem #progresseveryday #tech by @priyal.py

1k

9

14 hours ago

kvcache manager and scheduling #learningtogether #womeninstem #progresseveryday #tech #ai by @priyal.py

343

2

2 days ago

speculative decoding #learningtogether #womeninstem #progresseveryday #tech #consistency by @priyal.py

455

4

5 days ago

finetuning resources #learningtogether #womeninstem #progresseveryday #tech by @priyal.py

433

3

6 days ago

tensor parallelism in llm overview #learningtogether #progresseveryday #tech #consistency #womeninstem by @priyal.py

1k

20

7 days ago

llm serving #learningtogether #womeninstem #progresseveryday #tech #consistency by @priyal.py

384

3

10 days ago

1. Multi-Agent Research Assistant - Build multiple agents where one searches the web, one summarizes findings, one validates sources, and one generates the final report. Great for learning orchestration.Tech Stack: LangGraph / CrewAI, OpenAI, Tavily, FastAPI, Python2. SQL Agent over Database (Tool-Using Agent)Ask natural language questions like “top 5 customers by revenue” -> agent plans SQL query -> executes -> returns insights.Tech Stack: LangGraph, SQLAlchemy, PostgreSQL, OpenAI API, FastAPI3. Domain-Specific Chat Fine-Tuning (Fine-Tuning)Fine-tune an LLM on niche conversations (legal, healthcare, support) for more domain-accurate responses.Tech Stack: Llama/Mistral, HuggingFace, PEFT, LoRA, PyTorch4. Retrieval-Augmented Fine-Tuning (RAFT Style)Combine retrieval with lightweight fine-tuning to improve factual domain QA performance.Tech Stack: FAISS / ChromaDB, QLoRA, PyTorch, LlamaIndex5. LLM Evaluation Framework (LLMOps)Build a framework to measure hallucination, answer relevancy, faithfulness, latency, and cost across prompts/models.Tech Stack: RAGAS, DeepEval, LangSmith, Python6. AI Browser Automation Agent Agent browses websites, extracts info, fills forms, automates repetitive workflows.Tech Stack: Playwright, LangGraph, OpenAI API, Python7. Recommendation Agent (Retrieval + Ranking)User gives preferences -> agent retrieves candidates -> ranks -> returns personalized recommendations. Tech Stack: Embeddings, Vector DB, Python, FastAPI8. Self-Reflection Agent (Reasoning Loop)Agent critiques its own answer, identifies mistakes, retries, and improves output quality.Tech Stack: LangGraph, OpenAI API, Python9. Multi-Agent Financial Analyst (Agent Workflow)One agent extracts KPIs, one analyzes risks, one summarizes earnings reports, one compares quarters.Tech Stack: OCR, Pandas, LangGraph, OpenAI API, Python#learningtogether #progresseveryday #tech #consistency #ai by @priyal.py

472

9

11 days ago

slm with llm in a routing pipeline with LiteLLM for model routing with fallback logic, max retries etc. Hope this is helpful #learningtogether #womeninstem #progresseveryday #tech #consistency by @priyal.py

1k

28

13 days ago

1. LiteLLM – Lightweight open-source gateway for routing across OpenAI, Anthropic, Gemini, Azure, Bedrock, Ollama, etc.2. OpenRouter – One API endpoint for multiple LLM providers. Great for quick experiments and model switching.3. Helicone – Open-source LLM proxy with caching, logging, analytics, and observability.4. Portkey AI – AI gateway for retries, load balancing, fallbacks, caching, and provider routing.5. Bifrost - supports a broad range of providers 6. Kong AI Gateway – API gateway adapted for AI traffic, auth, security, and provider routing.7. Vercel AI Gateway – Unified gateway for multiple model providers with failover, observability, and OpenAI-compatible access.#learningtogether #womeninstem #progresseveryday #tech #consistency by @priyal.py

498

18

15 days ago

mixed precision overview #learningtogether #womeninstem #progresseveryday #machinelearning #tech by @priyal.py

348

2

16 days ago

1. Resume Screener - Uses VLMs to read resume PDFs/screenshots and rank candidates from visual + textual content.2. Invoice Processing - Extracts totals, vendor names, and tables from invoice images using document VLMs.3. Visual Customer Support Bot - Users upload product/device photos and the VLM diagnoses issues conversationally.4. Medical Scan Report Generator - Generates findings from X-rays or MRI scans using medical vision language models.5. E-commerce Product Tagger - Automatically creates product titles, descriptions, and attributes from product images.6. Multimodal Search Engine - Lets users search documents/images with natural language queries using CLIP embeddings.#learningtogether #womeninstem #progresseveryday #tech #consistency by @priyal.py

344

4

18 days ago