Piyush Makhija

CTO

San Francisco, California, United States10 yrs experience

AI EnabledAI ML Practitioner

Key Highlights

Led AI projects driving 20% margin improvements.
Architected multimodal AI systems for education.
Pioneered production-grade conversational AI solutions.

Stackforce AI infers this person is a SaaS and AI specialist focused on education and e-commerce solutions.

Contact

piyushmakhija5@gmail.com LinkedIn

Skills

Core Skills

Artificial IntelligenceProduct ManagementData ScienceNatural Language ProcessingDeep LearningMachine LearningComputer VisionTelecommunications

Other Skills

4G ProtocolsAIAI SystemsASRAlgorithmsAnalyticsBenchmarkingChatbot DesignCode mixedConvolutional Neural NetworksData AnalysisData IntelligenceData StructuresE-CommerceEvaluation Frameworks

About

I am an AI Architect & Product Builder specializing in creating durable, vertical AI systems and autonomous agentic workflows that drive measurable economic value. My career focus is on moving systems from mere "demo-ware" to reliable, high-scale production architectures. Core Expertise & Impact - Deep Tech Moonshots (Meraki Labs): Led AI for a canvas-first, multimodal tutor that operated on handwritten text and images. Solved critical real-time latency challenges by driving 20+ technical rebuilds to achieve production readiness, complete with reusable evaluation frameworks and autonomous video generation pipelines - Autonomous Agent Architecture: Architected and deployed the agentic data intelligence layer for a high-volume platform (processing $160M+ GMV). This system autonomously converts fragmented business data into real-time profitability insights, identifying opportunities for up to 20% margin improvement - Conversational AI & Scale: Built and optimized production-grade conversational agents (at Skit AI & Vahan) that successfully drove 2,000+ job placements monthly and automated critical business workflows, proving the measurable economic value of conversational AI My thesis: - AI should handle detail-oriented execution autonomously while humans focus on strategy - The difference between demos and durable products is architecture + evals + reliability engineering M.S. CS (Machine Learning) @ Georgia Tech | B.Tech @ IIT Roorkee 🌐 piyushmakhija.com | 💻 github.com/piyushmakhija5

Experience

Meraki labs

Head of Artificial Intelligence

Jan 2024 – Jan 2025 · 1 yr · Bengaluru

Meraki Labs is a venture studio building deep-tech "moonshots" to transform the world.
Building fermi.ai: Architecting the core intelligence for a tablet-first, multimodal AI tutor that replicates a world-class human teacher. The system perceives handwriting, images, and voice in real-time to provide Socratic "nudges", helping students learn how to solve problems rather than giving them answers.
Frontier Multimodal Engineering: Engineered a real-time voice & vision pipeline capable of processing handwritten inputs and spoken queries simultaneously. Achieved <1s voice latency and <20s complex reasoning time by iteratively re-architecting the stack 20+ times to optimize context windows and token generation speed.
Eval-Driven Development: Moved beyond "vibes-based" tuning by designing a 4-level evaluation framework. Systematically benchmarked models to drive IIT-JEE Advanced solvability from 20% to 75%.

AIMultimodal SystemsReal-time ProcessingEvaluation FrameworksArtificial IntelligenceProduct Management

Moneyflo

Founder and CTO

Jan 2023 – Jan 2025 · 2 yrs · Bengaluru · On-site

Platform Architecture & Scale: Designed and deployed the omni-channel Data Intelligence Platform for D2C/e-commerce brands, synthesizing strategic insights across 17.5M+ orders and $160M+ in GMV.
AI Analyst Layer: Designed and deployed the proprietary 'AI Analyst' agentic system that synthesizes fragmented logistics and sales data into real-time profitability insights, enabling clients to uncover and realize up to 20% margin improvement.
Technical Leadership: Led the entire Product, Architecture, and Engineering roadmap, focusing on system reliability, high data integrity, and establishing core evaluation frameworks for the intelligence layer.

Data IntelligenceE-CommerceAI SystemsProduct ManagementData Science

Skit.ai

ML Lead

Mar 2022 – Oct 2022 · 7 mos · Bengaluru · Hybrid

Pioneering Speech Architecture: Led 10+ engineers to architect and deploy production-grade, Pre-LLM English-Hindi ASR (wave2vec) and TTS (Tacotron-v2/FastPitch) pipelines. Achieved a sub-10% WER on mixed data, outperforming major cloud APIs.
Optimization and Cost Impact: Drove low-latency inference by implementing manual feature engineering and ONNX optimization. This architecture reduced external API reliance, saving over $10K per month.
New Revenue Streams: Monetized the proprietary TTS capability by launching tailored on-premise solutions, establishing a new B2B revenue stream that generated $8K in Monthly Recurring Revenue (MRR).

Speech ArchitectureASRTTSTechnical LeadershipNatural Language ProcessingDeep Learning

Vahan inc.

Machine Learning Lead

Apr 2018 – Oct 2021 · 3 yrs 6 mos · Bengaluru, Karnataka, India · Hybrid

Built a WhatsApp-based job screening chatbot (Python, FastAPI, Rasa) with 96%+ F1 on intent/entity recognition for noisy, code-mixed English–Hindi text.
Published 2 peer-reviewed papers on NLP for code-mixed and noisy text based on this work.
Developed ML systems for ID verification, calling prioritization, marketing cohorts, and churn prediction, improving tele-caller efficiency and campaign performance.
Helped shape a culture of collaborative learning through brown-bag sessions and leadership roundtables.

Chatbot DesignMachine LearningNatural Language ProcessingProduct Management

Rakuten

Graduate Research Intern

Aug 2017 – Nov 2017 · 3 mos · Tokyo, Japan · On-site

Real-Time Pedestrian Detection on Nvidia Jetson Tx2 from 30m+ distance using Convolutional Neural Network for Drone project and Proof of Concept for communication between on-flight and on-Ground systems using ROS

Convolutional Neural NetworksReal-Time ProcessingResearch and DevelopmentComputer Vision

Samsung electronics

Lead Engineer

Jun 2013 – May 2017 · 3 yrs 11 mos · Bengaluru Area, India · On-site

Designed and executed eMBMS, eMBMS Relay, and GCSE features in the 4G protocol stack, enhancing network efficiency and user experience.
Published a paper at IEEE WCNC 2017 and hold 4 granted US patents in wireless communications.
Led critical feature development and FPGA/chipset testing to launch a new chipset with robust pre-deployment performance.

Telecommunications4G ProtocolsResearch and Development