Sahil Jagtap

Companies I've built

HomePlate - CTO and co-founder. HomePlate is the DoorDash for senior nutrition: we turn idle licensed senior kitchens into on-demand medical meal prep for elderly patients who need medically-tailored food. Millions of seniors have specific dietary requirements driven by chronic illness, and the existing infrastructure doesn't handle medical complexity. We're solving that. $4,000 in revenue across the first 10 pilots. Picked up by Fox News. I turned down multiple cofounder, CTO, and founding engineer roles at other companies to go all-in here with Zach - that's how much I believe in this.

AgentBudget (demos) - creator and author. AgentBudget is the ulimit for AI agents: as autonomous agents proliferate, runaway cost accumulation is one of the most underappreciated failure modes in production AI. I built the circuit breaker. 3,000+ installs on PyPI, 100+ GitHub stars, and the project is now community-driven - contributors are shipping features independently. Published a technical whitepaper on Zenodo formalizing the architecture. Currently exploring integration with Coinbase's x402 autonomous payment protocol for agent-native billing.

Airstitch - co-founder and CTO. Your AI Chief of Staff living inside iMessage, SMS, and WhatsApp - no new app, no new interface, just intelligent orchestration over the texting surface everyone already uses. 200+ active users, 20+ paid within 9 days of launching beta. Built as part of Founders Inc Artifact'26. Acquired. The acquisition validated the core thesis: AI-native interfaces built on ambient surfaces people already live in are the next distribution channel.

Barty - founder. Barty was an AI-powered recruiter that automated candidate screening and interview workflows. Won $8,000 at George Mason's Patriot Pitch Competition (STEM track, 1st place). Barty didn't reach escape velocity. Every founder I respect has a failed company in their history. Barty taught me more about B2B sales, enterprise procurement friction, and the gap between "people love the demo" and "people pay for the product" than anything else - and that knowledge directly shaped how I built Airstitch.

Research and technical work

NSF Wearables Research - George Mason University. Lead ML engineer on a federally-funded research project at the intersection of wearable computing and cognitive health. The work focuses on early detection of cognitive impairment in neurodivergent individuals using spatial navigation patterns from wearable sensors. I architected the entire ML pipeline from sensor input to edge inference: a novel spatial-navigation model achieving 87% accuracy in preliminary validation studies, end-to-end data infrastructure processing 1.2M+ daily data points from 500+ participants, federated learning to preserve privacy across distributed datasets, and a 68% reduction in inference time via model optimization for real-time edge processing. Submitting to CHI 2026 as lead author. Speaking at the NARRTC Annual Conference 2026: "AI-Assisted Annotation Approach for Understanding Task Engagement of Neurodivergent Employees."

Cerebras Systems - Engineering Fellow. Selected for Cerebras' engineering fellowship - one of the most competitive technical programs in AI infrastructure. Built agentic AI systems on the world's fastest AI inference chip. Cerebras is building the hardware substrate that frontier model training depends on; being inside that as a fellow means working at the actual edge of what's possible in inference-time compute.

Microsoft - Student Researcher. Selected from 1,000+ applicants. Used Microsoft's Planetary Computer to design and deploy scalable ML pipelines over petabytes of Sentinel-1 radar satellite data - unlocking real-time environmental monitoring for climate research that previously required months of batch processing. The kind of infrastructure work that doesn't get press but moves fields.

Trigent Solutions - Software Engineer. Led AI/ML initiatives for FDA-regulated healthcare clients. Built an AWS Lambda-based computer vision system that cut medical imaging analysis time by 64% - presented directly to FDA representatives. Engineered a HIPAA-compliant data pipeline handling 200K+ patient records. Anomaly detection system improved fraud detection rates by 27%. All of this in a 4-month internship.

Zepto - Delivery Operations Research. I did delivery runs for Zepto - India's fastest-growing quick-commerce company (YC-backed, 10-minute grocery delivery) - because I wanted to understand last-mile logistics from first principles, not from a dashboard. Delivered groceries to my own neighbors. Found 3 critical friction points costing 15+ minutes per delivery at a company processing 100K+ orders daily. This is how I think about problems: go do the thing, then figure out what's broken.

Production rescue - NYC startup. Diagnosed and fixed a critical production bug blocking an early-stage NYC startup's entire revenue pipeline. They went from near-bankruptcy to $1M ARR within 6 months of the fix shipping. A single well-placed technical intervention can be worth more than months of product work.

Open source

AgentBudget - 3,000+ PyPI installs, 100+ GitHub stars, now community-owned. Production contributions to LangChain, Supabase, LangFuse, and Scribe - tools used daily by thousands of developers building AI applications. Built a real-time swimmer position estimation model using RNN-LSTM on sensor data with no GPS and no cameras - won at hackathon. (Devpost)

Speaking, judging, and mentorship

Speaker, NARRTC Annual Conference 2026 - presenting original AI + accessibility research to a national audience of rehabilitation and neurodevelopmental practitioners.

Judge, NASA TechRise - evaluated student aerospace and technology projects for one of NASA's flagship innovation programs.

Judge, Technica - the world's largest hackathon for underrepresented genders. Evaluated AI and systems projects across hundreds of submissions.

Judge, UVA HooHacks and UMD Bitcamp - two of the most competitive university hackathons on the East Coast.

Mentor in Residence, The MIX Incubator (GMU) - advising 20+ early-stage founders on product, fundraising, and go-to-market. Credibility here comes from actually winning $8K at Patriot Pitch and building two companies through the same environment.

Awards and recognition

HackMIT (MIT), PennApps (UPenn, 2023 & 2024), HackPrinceton (Princeton) - wins at three of the most competitive university hackathons in the country. 30+ total events attended.

Patriot Pitch 1st Place, STEM Track - $8,000 award at George Mason's premier entrepreneurship competition, beating every other STEM-category team.

Mason Distinction Award - $84,000 scholarship for exceptional academic and leadership potential. Competitive, not automatic.

Cerebras Engineering Fellowship - one of the most selective technical programs in AI infrastructure.

Microsoft Student Researcher - selected from 1,000+ applicants nationally.

Demo Day, ProductFest (RVATech) - presented to Virginia's largest product and technology community.

Programs and community

Founders Inc, Artifact'26 Batch - one of SF's most selective pre-seed founder programs. Alums include LiveKit and Lunabill (YC F25).

NVIDIA Inception - NVIDIA's program for cutting-edge AI and data science startups.

Former President, Data Science & Computation Society (GMU) - scaled membership from fewer than 30 to 500+ students. Built partnerships with Columbia University's NSDC, Google, Amazon, and Microsoft. Ran 50+ technical events and workshops. Left it meaningfully larger than I found it.

Member, Noisebridge (SF) and MIX Makerspace (Fairfax, VA) - both active working spaces I actually use.

CS @ George Mason University · GitHub · LinkedIn · X · Email · AgentBudget

Side experiments

Poop - tried to create a clothing brand. Sometimes you just have to see what happens.