Freelancers using LangChain in MumbaiFreelancers using LangChain in Mumbai
ProjectsPeople
Anurag Nagare's avatar
Anurag Nagare
Mumbai, India
I’m an AI & Machine Learning engineer with expertise in deve
I’m an AI & Machine Learning engineer with expertise in deve
Cover image for Most AI research tools are
Most AI research tools are just a chatbot with a search button. I built something different. Every time you ask an AI to research something, you're getting one model, one pass, no quality check. It writes confidently, cites poorly, and you have no idea if what it produced is actually accurate. For anyone making real decisions from AI-generated research, that's a silent risk most people ignore. The problem gets worse at scale the longer and more complex the question, the more a single model hallucinates, misses sources, and loses structure. There's no one checking its work. So I built ResearchOS a 5-agent pipeline where each agent has one job. A Supervisor breaks down your question. A Researcher runs parallel searches across 22+ sources. An Analyst extracts data and auto-generates charts. A Writer synthesises a cited report. A Critic fact-checks it and sends it back for revision if anything is wrong. The loop runs up to 3 times before the report is approved. One question in. A full cited report with charts and PDF export in under 10 minutes. I tested it live by watching the Critic catch a missing citation mid-run and send the Writer back to fix it before approval. That's the part that makes this actually usable for real work. Built on LangGraph, Groq, Tavily, ChromaDB and runs entirely on free tiers.
1
79
Cover image for HybridAlpha (Hybrid RAG) : One tool
HybridAlpha (Hybrid RAG) : One tool digs into actual SEC filings, not just static documents. From EDGAR, it grabs 10, Ks and 10, Qs fresh each time. Sections like MD&A or Risk Factors get split out by name during parsing. Storage happens two ways at once: words go to ChromaDB, numbers land in SQLite. When a question arrives, the router decides, tone, driven, number, heavy, or both. Depending on that choice, the query moves to one place, sometimes both. Context flows forward only after sorting is done. Answers come from Llama 3.3 70B via Groq, always tagged with sources. Each output ties back to where the data lived. Start by asking, What risks did Apple highlight regarding AI rivals? Out comes exact quotes pulled straight from official documents.
0
49
Cover image for I created WealthWise Agent, a smart personal finance planner...
I created WealthWise Agent, a smart personal finance planner designed to craft personalized budget plans and investment strategies. This app takes into account user inputs like salary, expenses, and financial goals, and then uses a Large Language Model (Gemini) to analyze these factors based on the 50/30/20 budgeting rule. It offers a clear step-by-step reasoning log, a detailed JSON-structured financial plan, and an interactive visualization of budget allocation, empowering users to make informed choices to reach their financial goals. 💻 Tech Stack Used: Frontend/UI: Gradio (custom themed with CSS, Orbitron font) AI/Logic: Google Gemini (gemini-1.5-flash) with LangChain agents Data: yFinance API for real-time stock/ETF data, Pandas & NumPy for calculations Visualization: Plotly Express for interactive charts
0
51
Cover image for What your attention heatmap isn't
What your attention heatmap isn't telling you Everyone's staring at attention heatmaps and calling it "interpretability." Almost nobody's asking whether a single attention map actually tells you what the model used to make its decision. It doesn't. Not on its own. A raw attention map from layer 8 shows you what layer 8 attended to. It says nothing about how that signal got mixed, diluted, or overwritten by every layer before and after it. Attention rollout fixes this — and I built a walkthrough to show why it matters. Here's what makes it more than a "pretty heatmap" demo: Instead of visualizing one layer's attention, I traced how information actually flows through the full transformer stack. → Every layer's attention matrix is extracted, per head, per token → Multi-head attention is averaged, then combined with the residual connection (identity + attention) — this is the step most tutorials skip, and it's the one that actually matters → The combined matrices are matrix-multiplied layer by layer, rolling attention forward from input to output → The result: a single map showing genuine token-to-token influence across the entire network, not just one layer's snapshot The overlay shows you everything: → Per-layer attention vs. rolled-out attention, side by side → Token importance scores overlaid directly on the input text → A comparison view: which tokens raw attention says "matter" vs. which ones rollout says actually matter → Head-level breakdown so you can see which heads specialize vs. which are noise No black box. No "trust me, the model looked here." Just linear algebra, applied honestly across every layer instead of cherry-picking one. Built with PyTorch + HuggingFace Transformers + Matplotlib. Runs on any pretrained transformer, fully offline. ⚠️ Important: attention rollout is an approximation, not ground truth. It assumes attention is the primary information pathway, which ignores MLP layers and can still mislead for very deep models. Treat it as a debugging lens, not proof of causality.
0
26
LangChain(3)
Amol Bhosale's avatar
Amol Bhosale
Mumbai, India
I build custom SaaS platforms & MVPs (React, Node, Next.js
I build custom SaaS platforms & MVPs (React, Node, Next.js
Cover image for AI-Powered Energy Optimization System Development
AI-Powered Energy Optimization System Development
0
7
Cover image for GDPR AI-Powered Appointment Scheduling Platform
GDPR AI-Powered Appointment Scheduling Platform
0
13
Cover image for ISO & GDPR Carbon Accounting & Energy Management Platform
ISO & GDPR Carbon Accounting & Energy Management Platform
0
35
Cover image for Internal Ops Tool for accountancy Firm
Internal Ops Tool for accountancy Firm
0
13
LangChain(1)
Kushal Harsora's avatar
Kushal Harsora
Mumbai, India
A web developer who builds responsive and modern websites.
A web developer who builds responsive and modern websites.
Cover image for TextyBit - A Gemini based RAG Application for your PDF files
TextyBit - A Gemini based RAG Application for your PDF files
0
0
Cover image for Shree Ambika Enterprises
Shree Ambika Enterprises
1
0
Cover image for Avinnya Skin Clinic
Avinnya Skin Clinic
1
1
Placeholder project card media
View more →
LangChain(1)