--- title: "Aradhya Khandelwal - Co-Founder in India | Tal" description: "Meet Aradhya Khandelwal, Machine Learning Engineer | Audio & Signal Processing, Edge AI (PyTorch, CUDA, Jetson) | LLM Agents | IEEE InGARSS ’25 based in India. AI stack includes anthropic apis, antigravity, and claude code. Explore public work links including github and aradhya-khandelwal.vercel.ap..." image: "https://tal.club/meet/aradhya-khandelwal/share-image" --- # Aradhya Khandelwal on Tal ## Headline Machine Learning Engineer \| Audio & Signal Processing, Edge AI \(PyTorch, CUDA, Jetson\) \| LLM Agents \| IEEE InGARSS ’25 ## About I'm an ML engineer in Bengaluru\. Mostly audio and signal processing, models that run on small hardware, and a bit of LLM agent stuff on the side\. Right now I'm at SSS Defence, where I've spent the last few months teaching a mic array to hear drones\. The fun part: there wasn't enough real data, so I trained the whole thing on synthetic audio I generated myself\. It still caught 89\.4% of drones across 40 real field flights, even when 15 to 30 of the mics were dead\. I also wrote the CUDA front end \(GCC\-PHAT, beamforming\), got it running on a Jetson Orin at 1\.5x real time, and spent one very long week chasing a memory leak that only showed up after an hour of running\. Outside work, I co\-founded The Algorithm Lab\. It's a small AI marketing startup with paying clients, and I'm the only engineer, so I built the whole platform: a bunch of LLM agents that plan and write campaigns, a critic agent that tears them apart, and a website audit tool\. FastAPI, Postgres, and way too many tests \(594\)\. I also built Hypercharge, an open\-source tool for coding agents, because I got tired of Claude editing files it never actually opened\. It gives the agent a map of the codebase, memory between chats, and checks so it can't say "fixed" while the tests are still failing\. Currently messing around with Earmark, a small C\+\+/WASM voice isolation experiment\. No results yet, ask me in a month\. Before all this I spent 9 months at Intact India writing Python automation for 25\+ regression workflows, which cut manual QA time by about 40%\. Not glamorous, but it's where I started loving tests\. And I did research at IIT Roorkee on radar \(PolSAR\) data, which turned into a paper at IEEE InGARSS 2025, plus another one at Springer ICDSA 2025\. In college I ran Techstars Startup Weekend Jaipur with a team of 150 students\. Taught me more about shipping under pressure than any course did\. Looking for ML roles in audio, speech, signal processing or edge AI, anywhere the model has to actually work in the real world, not just in a notebook\. arukhandelwal281@gmail\.com aradhya\-khandelwal\.vercel\.app github\.com/arukh281 ## Location India ## AI stack anthropic apis antigravity claude code hugging face langgraph ollama ## Work experience Co\-Founder at The Algorithm Lab, Jun 2026 \- Present · 5 mos, Only engineer at an AI marketing startup with 6 clients paying $1\.5K a month in total for a managed service\. Building the product that automates that service, now in beta\. Built 7 platform\-writer agents, a topic planner, an adversarial critic and a 4\-stage campaign planner, all on one company audit\. Built the website\-audit pipeline: HTML extraction, headless Playwright screenshots, Claude vision grading and PDF reports\. Designed a provider layer on OpenRouter, with Claude as the primary model and Gemini as a per\-workload fallback\. FastAPI, async SQLAlchemy with PostgreSQL and React 19 on Render; 594 automated tests\. Deep Learning Specialist at SSS DEFENCE, Mar 2026 \- Present · 8 mos, Built a deep\-learning acoustic drone detector for a multi\-microphone array, trained only on synthetic data\. On 40 real field flights it caught 89\.4% of drone segments \(95% CI 87\.7–90\.9%\), even with 15–30 dead microphones; synthetic AUROC 0\.995\. Built the synthetic data generation pipeline: multirotor acoustic synthesis, per\-microphone array realism \(gain, jitter, dead channels\), and time\-aligned labels\. Diagnosed the model with branch knock\-out ablations, and added synthetic machinery hard negatives that cut look\-alike scores from 0\.79 to 0\.04–0\.16 on synthetic tests\. Built and debugged a CUDA signal\-processing front\-end \(whitening, cross\-spectra, GCC\-PHAT, beamforming\), ported the detector to NVIDIA Jetson Orin with runtime tests and benchmarks, and built a Raspberry Pi field\-recording rig\. Automation Specialist at Intact India, Jul 2025 \- Mar 2026 · 9 mos, Built Playwright/Python automation for 25\+ regression workflows, cutting manual QA time by ∼40%, and wired automated validation into CI/CD during the consolidation of 3 subsidiary firms\. Research Intern at Indian Institute of Technology, Roorkee, Jun 2025 \- Jul 2025 · 2 mos, Roorkee, Uttarakhand, India, During my internship at IIT Roorkee, I worked on developing a deep learning\-based unsupervised approach to classify volume scattering for analyzing surface roughness and potential water\-ice deposits\. Excited to share that this work has now been accepted for presentation at IEEE InGARSS 2025\! Convenor at Techstars Startup Weekend Jaipur, Apr 2024 \- Nov 2024 · 8 mos, Jaipur, Rajasthan, India, Was closely involved in managing the event from start to finish, working with and coordinating a team of 150 students\. Together, we hosted a three\-day event that brought people together and helped lay the foundation for a growing startup community\. ## Education Bachelor of Technology, Computational Science, Manipal University Jaipur 12th, PCM, the Espee global school ## Public links - github: https://github\.com/arukh281 - aradhya\-khandelwal\.vercel\.app: https://aradhya\-khandelwal\.vercel\.app/ [View the original profile](https://tal.club/meet/aradhya-khandelwal) ```json { "@context": "https://schema.org", "@graph": [ { "@type": "Person", "@id": "https://tal.club/meet/aradhya-khandelwal#person", "name": "Aradhya Khandelwal", "url": "https://tal.club/meet/aradhya-khandelwal", "image": "https://tal.club/meet/aradhya-khandelwal/profile-photo.jpg", "description": "Meet Aradhya Khandelwal, Machine Learning Engineer | Audio & Signal Processing, Edge AI (PyTorch, CUDA, Jetson) | LLM Agents | IEEE InGARSS ’25 based in India. AI stack includes anthropic apis, antigravity, and claude code. Explore public work links including github and aradhya-khandelwal.vercel.ap...", "jobTitle": "Co-Founder", "worksFor": { "@type": "Organization", "name": "The Algorithm Lab" }, "hasOccupation": { "@type": "Occupation", "name": "Co-Founder" }, "homeLocation": { "@type": "Place", "name": "India" }, "alumniOf": [ { "@type": "EducationalOrganization", "name": "Manipal University Jaipur" }, { "@type": "EducationalOrganization", "name": "the Espee global school" } ], "knowsAbout": [ "anthropic apis", "antigravity", "claude code", "hugging face", "langgraph", "ollama" ], "sameAs": [ "https://github.com/arukh281", "https://aradhya-khandelwal.vercel.app/" ], "mainEntityOfPage": { "@id": "https://tal.club/meet/aradhya-khandelwal#webpage" } }, { "@type": "ProfilePage", "@id": "https://tal.club/meet/aradhya-khandelwal#webpage", "url": "https://tal.club/meet/aradhya-khandelwal", "name": "Aradhya Khandelwal on Tal", "description": "Meet Aradhya Khandelwal, Machine Learning Engineer | Audio & Signal Processing, Edge AI (PyTorch, CUDA, Jetson) | LLM Agents | IEEE InGARSS ’25 based in India. AI stack includes anthropic apis, antigravity, and claude code. Explore public work links including github and aradhya-khandelwal.vercel.ap...", "dateModified": "2026-10-02T05:57:26.382Z", "isPartOf": { "@id": "https://tal.club/#website" }, "mainEntity": { "@id": "https://tal.club/meet/aradhya-khandelwal#person" }, "primaryImageOfPage": { "@type": "ImageObject", "url": "https://tal.club/meet/aradhya-khandelwal/profile-photo.jpg", "contentUrl": "https://tal.club/meet/aradhya-khandelwal/profile-photo.jpg", "caption": "Aradhya Khandelwal profile photo on Tal" }, "inLanguage": "en-IN" } ] } ```