Harsh Saxena

AI Researcher

Noida, Uttar Pradesh, India6 yrs 1 mo experience

Highly StableAI Enabled

Key Highlights

Boosted recommendation accuracy by 35% with BeautyGPT.
Engineered GPT-4 conversational systems with semantic search.
Mentored teams, improving code quality and release velocity by 40%.

Stackforce AI infers this person is a B2C AI solutions architect with expertise in generative AI and computer vision.

Contact

Skills

Core Skills

Generative AiComputer VisionMlopsMachine Learning

Other Skills

Artificial Neural NetworksAutoencodersC++CNNConvolutional Neural Networks (CNN)Data AnalysisData ScienceDeep LearningEngineeringFacial RecognitionFastAPIFeature EngineeringGPT-4Generative Adversarial Networks (GANs)Generative Neural Networks

About

I help companies turn cutting-edge AI into customer-facing products.With 5 years of experience building scalable AI systems, I specialize in: 🔹 Generative AI (LLM, GPT-4, LangChain, RAG, Weaviate) 🔹 Computer Vision (CNNs, Transformers, Image generation) 🔹 MLOps (C++, FastAPI, Docker, mlflow, AWS) At Orbo.ai, I led the creation of BeautyGPT — a multimodal GenAI system that boosted recommendation accuracy by 35%. I’ve also engineered GPT-4 powered conversational systems with sematic search. Whether it's developing face verification for KYC, facial landmarks for AR applications, coding inference in C++ to achieve 10× speedups or mentoring teams — I focus on delivering performance, impact, and speed. 💡 If you're working on GenAI, vision systems, or scaling AI to millions — let’s connect.

Experience

6 yrs 1 mo

Total Experience

5 yrs 2 mos

Average Tenure

11 mos

Current Experience

Fractal

Senior Computer vision | Gen AI Engineer

Jul 2025 – Present · 11 mos · Noida, Uttar Pradesh, India · Hybrid

Vision Language Data Scientist

Orbo.ai

3 roles

Senior Computer Vision Engineer | Gen AI & ML Solutions Architect

Promoted

Apr 2024 – Jul 2025 · 1 yr 3 mos

> Spearheaded design and deployment of Generative AI-powered conversational systems using GPT 4o, LangChain, and vector search partnering with product owners to boost response relevance by 35%.
>Launched BeautyGPT, a multimodal RAG recommendation engine integrating computer-vision diagnostics with GPT-4 chain-of-thought reasoning driving a 35% uplift in recommendation accuracy
>Mentored and led a team of 5 engineers, establishing a feedback-driven culture and improving code-quality and release velocity by 40% through rigorous testing and reviews

Generative AIGPT-4LangChainvector searchrecommendation accuracyComputer Vision

Computer Vision Engineer | ML Solutions Architect

Jul 2020 – Mar 2024 · 3 yrs 8 mos

> Designed and fine-tuned custom CNN and Transformer models for AR makeup serving 500K+ users with 15ms latency.
>Architected C++-based ONNX inference pipelines and OpenCV optimizations for KYC face matching achieving a 10× latency reduction to support 24/7 real-time operations.
>Authored and standardized 20+ API documentation sets, model cards, and internal data dictionaries reducing onboarding time by 30% and accelerating cross-team collaboration
>Developed and enforced QA processes and model-governance standards aligning with enterprise compliance frameworks and decreasing post-deployment defects by 25%

CNNTransformer modelsC++ONNXOpenCVComputer Vision+1

Computer Vision Intern

Mar 2020 – Jun 2020 · 3 mos

Tdb.ai

Computer Vision Developer

Oct 2019 – Mar 2020 · 5 mos · Indore, Madhya Pradesh · Remote

>Collaborated directly with the CEO to translate 5 client visions into fully functional computer-vision prototypes, delivering each solution on schedule and increasing repeat business inquiries by 30%
>Mentored and managed a team of 3 computer-vision interns, designing a streamlined onboarding program that cut ramp-up time in half and boosted intern contribution by 50%
>Standardized development workflows and code-review processes, reducing post-launch defects by 25% and accelerating project iteration cycles by 20%

Computer VisionPythonFeature EngineeringMachine Learning

Indian institute of technology, kanpur

Computer Vision Research Internship

Jun 2019 – Jul 2019 · 1 mo · Kanpur Area, India

>Collaborated with Ph.D. researchers to design and implement a stereo-camera calibration and depth-estimation pipeline, boosting depth-map accuracy by 20% while slashing calibration time by 30%
>Developed a subject-localization algorithm using paired stereo images.
>Authored detailed validation tests and documentation for calibration workflows, standardizing experiments and cutting iteration cycles by 40%

Stereo-camera calibrationdepth-estimation pipelineComputer Vision

Defence research and development organisation (drdo)

2 roles

Computer Vision Research Intern

Jan 2019 – Jan 2019 · 0 mo · Dehradun, Uttarakhand, India

Skills Learned - Image Classification, face recognition, CNN, ANN, multi target tracking.
Developed a computer vision method to stabilize the image sequence recorded by a panning camera. Also published a research paper on the same in IEEE. Also learnt deep learning (Image Classification, Neural style transfer, Face Recognition, YOLO detection).

Image Classificationface recognitionComputer Vision

Computer Vision Research Internship

Jun 2018 – Aug 2018 · 2 mos · Dehra Dun Area, India

Skills Learned - Image Segmentation, Object detection and tracking, Image processing, Video stabilization.
Developed two projects one titled "Multi target detection and tracking with Video Stabilization" and another titled "Intelligent Auto Visual Tracker" and also published a research paper on second project in springer international.

Image SegmentationObject detectionComputer Vision