Freelancers using Python in Thane

Freelancers using Python in ThaneFreelancers using Python in Thane

Projects People

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 Everyone's building AR filters and

Everyone's building AR filters and calling it "computer vision magic." Almost nobody's asking what's actually happening underneath — that most of these effects are just clever masking, not detection. Here's proof. I built an invisibility cloak that runs entirely in the browser, no green screen, no chroma key, no model training. https://github.com/AnuragNagare/Ghost-frame

4

2

210

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

69

Cover image for Everyone's racing to add biometrics

Everyone's racing to add biometrics to logins. Almost nobody's asking what happens when you can't — or shouldn't — touch the sensor. Shared kiosks, clinical settings, accessibility needs, hygiene-sensitive environments. Fingerprint readers and face unlock assume contact or a stored faceprint. Sometimes you want authentication that touches nothing and stores no biometric image of you at all. So I built GestureAuth — a contactless authentication system where your "password" is a sequence of hand gestures performed in front of a standard webcam.

1

63

Cover image for Everyone's talking about AI in

Everyone's talking about AI in healthcare. Nobody's building low-cost tools for the people who actually need early answers. Neurologists are overbooked. Clinical tremor assessments require in-person visits, specialist equipment, and months of waiting. The 10 million people living with Parkinson's globally and the millions more who don't yet know have no accessible way to flag early symptoms from home. So I built TremorLens a real-time hand tremor detection tool that runs entirely on a standard webcam. Here's what makes it more than just a webcam project: Instead of simple motion detection, I built a full signal processing pipeline on top of computer vision. MediaPipe tracks 21 hand landmarks per frame. The index fingertip's x/y displacement is buffered across a 3-second rolling window. scipy FFT then decomposes that signal into its frequency components and flags dominant activity in the 4–6 Hz range clinically associated with Parkinson's resting tremors. The live overlay shows you everything: → Real-time FFT power spectrum with the tremor zone highlighted → Dominant frequency readout in Hz with a 10-frame rolling average for stability → Color-coded STABLE / TREMOR DETECTED indicator → Fingertip displacement graph and movement trail → Auto-saved CSV session log timestamp, frequency, amplitude, tremor flag every session

0

55

Abu Aasif Ansari

Bhiwandi, India

I build AI agents & internal tools that act on data

13: Followers

I build AI agents & internal tools that act on data

Cover image for Ask My Docs — RAG-Based AI Chat Agent

Ask My Docs — RAG-Based AI Chat Agent

0

1

Cover image for Sounds Like AI? — AI Writing Authenticity Checker

Sounds Like AI? — AI Writing Authenticity Checker

0

1

Cover image for Project Overview This is a personal portfolio and services s...

Project Overview This is a personal portfolio and services site built for a working freelancer — myself — as part of the Make It Real Challenge. I'm a data visualization specialist and AI dashboard developer, and I wanted the site itself to reflect that work rather than just describe it. So instead of stock photography and generic copy, every visual on the site is a real screenshot from a project I've actually shipped. The Concept Most freelance portfolio sites lean on polished lifestyle photography — laptops on wooden desks, people in blazers looking thoughtfully at monitors. I wanted to flip that: let the actual product screens do the talking. The site is built around a simple idea — "solutions built on data, not guesswork" — and that shows up literally in the visuals, not just the copy. What's on the site Homepage: A direct hero statement ("I Turn Messy Data Into Decisions") backed immediately by a real dashboard screenshot from my Anomaly Review & Action Console, followed by a "recent work" section pulling in three separate real projects. Services page: Four core offerings — Anomaly Detection Dashboards, Power BI & Data Analytics, AI-Integrated Internal Tools, and Custom Data Automation — each paired with an actual screenshot of that specific project (not a mockup). About page: A short, direct bio and a real photo, no filler. Contact page: A simple inquiry form paired with a data-visualization graphic that matches the site's overall theme. How I used Finish Layer Block Animations: I used on-appear and on-scroll triggers (Reveal and Slide styles) across the homepage, services, and contact pages. Section headings reveal as the page loads, and content blocks slide/fade in as the visitor scrolls — this turns what would be a static, all-at-once page into a guided, paced experience. Block Transform: Rather than a flat grid, I applied a subtle rotation to a couple of key visuals (a dashboard screenshot and the contact-page graphic) — just a few degrees, enough to break the rigidity of the layout without sacrificing readability. It gives the page an asymmetrical, more intentional feel instead of looking templated. Process I built this using Squarespace's Blueprint AI to get a fast base structure, then went through every page replacing AI-generated placeholder content (stock imagery, generic service descriptions, an accidental product/ecommerce section) with real project data, real screenshots, and copy that actually reflects how I work with clients. The site is password-protected rather than published on a paid plan, and is fully responsive across desktop and mobile. Access: 🔗 Site: https://pike-megalodon-4yb2.squarespace.com (https://pike-megalodon-4yb2.squarespace.com)🔑 Password: Omega

2

321

Cover image for Anomaly Review & Action Console (Retool + AI)

Anomaly Review & Action Console (Retool + AI)

1

6

Tanish Khandelwal

Data Guru. AI Solutions. ML Engineer.

Data Guru. AI Solutions. ML Engineer.

Cover image for tknishh/payclosur

tknishh/payclosur

0

17

Cover image for tknishh/uber-mage-data-analysis

tknishh/uber-mage-data-analysis

0

34

Cover image for tknishh/FileWise

tknishh/FileWise

0

17

Placeholder project card media

View more →

TechAPeek Services

🔍 Transforming Data into Business Insights

🔍 Transforming Data into Business Insights

Cover image for Dynamic Advertisement Overlay

Dynamic Advertisement Overlay

0

13

Cover image for Billing Management System

Billing Management System

0

7

Cover image for HUMOUR DETECTION USING NLP

HUMOUR DETECTION USING NLP

0

76

Placeholder project card media

View more →

FastAPI, AI & RAG Engineer helping startups ship faster

New to Contra

FastAPI, AI & RAG Engineer helping startups ship faster

Cover image for Shadow Watch — Behavioral Intelligence

Shadow Watch — Behavioral Intelligence & Account Security Platform Description Shadow Watch is a behavioral intelligence platform that continuously analyzes user interaction patterns to build behavioral profiles, detect anomalies, and strengthen account security without introducing friction to the user experience. Instead of relying solely on passwords or one-time verification checks, Shadow Watch observes behavioral continuity over time to establish trust signals and identify potentially compromised sessions. The system combines behavioral fingerprinting, activity analysis, risk scoring, and anomaly detection to help organizations recognize suspicious account activity before it escalates into an account takeover event. Key Features Behavioral fingerprint generation User activity intelligence Trust score calculation Session anomaly detection Account takeover prevention Temporal behavior analysis Privacy-conscious architecture Tech Stack Python • FastAPI • PostgreSQL • Security Engineering • Behavioral Analytics • Risk Scoring

0

23

Cover image for Grid CLI is a terminal-first

Grid CLI is a terminal-first developer platform designed to streamline repository management, project automation, diagnostics, and AI-assisted workflows through a unified command interface. Built around a custom shell architecture, Grid CLI enables developers to initialize projects, audit repositories, execute automated tasks, manage environments, and interact with AI-powered development tooling without leaving the terminal. The project focuses heavily on developer experience, modular command design, extensibility, and productivity automation. Key Features Custom interactive shell Repository diagnostics and auditing Project initialization workflows Automation and task execution Modular command architecture AI-assisted developer operations Cross-project workspace management Tech Stack Python • CLI Engineering • Automation • Git • Developer Tooling • Terminal UI

0

24

Cover image for Quant Terminal is a portfolio

Quant Terminal is a portfolio analytics and asset management platform designed to help investors monitor holdings, analyse diversification, and evaluate portfolio allocation strategies. The platform provides a real-time view of portfolio performance while offering allocation analysis across multiple asset classes including equities, commodities, cryptocurrencies, and foreign exchange markets. Key Features: • Portfolio holdings dashboard • Real-time P&L monitoring • Diversification analysis • Portfolio drift detection • Strategy-based portfolio views • Asset allocation management • CSV export functionality • Interactive desktop interface Tech Stack: Python, Data Analytics, Portfolio Modelling, Financial Data Processing, Desktop Application Development Role: Sole Developer and System Designer Outcome: Built a functional portfolio management terminal capable of tracking holdings, visualising allocation drift, and supporting portfolio rebalancing decisions through quantitative analysis.

0

28

Cover image for Anchor is an AI governance

Anchor is an AI governance platform designed to provide deterministic policy enforcement for AI systems and autonomous agents. The platform focuses on a critical problem facing enterprises adopting AI: ensuring that model-driven decisions remain explainable, auditable, and compliant with organizational policies. Key capabilities include: • Constitutional governance architecture • Policy-based execution controls • Deterministic decision validation • Replayable audit trails • Entity visibility and access governance • Runtime compliance verification • Governance observability and traceability The system is designed around the principle that governance decisions should be independently verifiable rather than dependent on opaque model behaviour. Tech Stack: Python, FastAPI, PostgreSQL, Docker, JWT Authentication, Audit Logging, Governance Engine Design Role: Founder, System Architect, Backend Engineer Project Links: Website: https://www.animuslab.dev (https://www.animuslab.dev)GitHub: https://github.com/AnimusLab/Anchor Research: https://zenodo.org/records/19734724

0

31

Sanchet Nagarnaik

Versatile full-stack developer exploring Blockchain and AI.

5.0: Rating

Versatile full-stack developer exploring Blockchain and AI.

Cover image for Chess Delay

0

10

Cover image for PhilTor - Phil Dunphy inspired Realtor mobile app

PhilTor - Phil Dunphy inspired Realtor mobile app

0

13

Cover image for MongoDB Masking

MongoDB Masking

0

9

Cover image for Social Media Integration

Social Media Integration

0

11

Technical Virtual Assistant & Web Operations

New to Contra

Technical Virtual Assistant & Web Operations

Cover image for AI Lead Generation Workflow

AI Lead Generation Workflow

0

2

Cover image for SEO Automation Workflow for Live Keyword Mining

SEO Automation Workflow for Live Keyword Mining

0

1

Cover image for Developed an application designed to

Developed an application designed to capture and analyze niche, long-tail conversational keywords frequently utilized within AI interfaces. Because traditional SEO tools often overlook conversational search intent phrases, this application maps and logs specific semantic variations into structured database layouts, enabling digital creators and businesses to optimize their content strategies for next-generation search behaviors.

0

43

Cover image for Designed and structured a comprehensive

Designed and structured a comprehensive comparative matrix detailing the unique capabilities of leading large language models (ChatGPT, Gemini, and Claude). This asset translates complex technical model parameters—such as context capacities, automation layers, and native execution engines—into highly scannable, engaging visual documentation tailored for digital platforms and tech-focused audiences.

0

35

Srikanth Tangirala

Data Analyst / Data Scientist

13: Followers

Data Analyst / Data Scientist

Cover image for Sentiment Analysis of Google Playstore App Reviews

Sentiment Analysis of Google Playstore App Reviews

0

42

Cover image for International Debt Statistics Analysis

International Debt Statistics Analysis

0

36

Cover image for English Premier League Table

English Premier League Table

0

19

Placeholder project card media

View more →