Pranav's Profound AI Rep
Pranav Pandey
Edges
Building Production AI Systems
Pranav owns the full stack of AI system delivery, from architectural design through data pipelines, evaluation infrastructure, and production deployment. He has built multi-agent systems, ingestion pipelines, and fine-tuning workflows from scratch across multiple companies. His instinct is to design for the whole system, not just the component in front of him.
Spotted in 3 Stories
A
Building AI Evaluation Infrastructure
Pranav builds the measurement systems that tell a team whether their AI is actually working. At SpotDraft, he built an evaluation harness from scratch that surfaced real customer failures, tracked regressions across prompt and tool changes, and automated prompt optimization based on failing cases. He designs eval systems to be reusable across teams, not one-off scripts.
Spotted in 2 Stories
ATranslating Research to Production
Pranav evaluates research papers against production constraints and makes deliberate calls on what to implement and what to skip. He looks for peer validation, available code, and fit with what the team can actually ship. At SpotDraft, he implemented a virtual file system architecture for agents from a paper because it reduced agent context load; he deferred fine-tuning papers because the data collection effort did not match the team's capacity at the time.
Spotted in 2 Stories
W
Training Deep Learning Models at Scale
Pranav has trained deep learning models across computer vision, audio, and NLP on large-scale datasets spanning millions of images, videos, and terabytes of text. He has worked with CNNs, RNNs, LSTMs, autoencoders, BERT, and generative architectures including CycleGAN, and has deployed models to mobile, browser, and cloud environments. This breadth of model training experience underpins his applied AI work.
Spotted in 1 Story
Pranav Pandey
About Me
Pranav Pandey is a Senior Applied AI Engineer who has spent the last five years building production AI systems across deep learning, computer vision, and large language models. Most recently at SpotDraft, he owned the full stack of a multi-agent AI platform for in-house legal teams, from architectural design through evaluation infrastructure and data ingestion pipelines, serving more than 400 enterprise clients.
Pranav's career spans a range of applied AI contexts. At Wipro, he worked on computer vision research including floor-plan recognition for a government authority in Dubai and a COVID detection system from cough audio. At XTRA, he built 3D human pose estimation and activity recognition systems deployed both on cloud and in-browser. At Motive, he worked on object detection, tracking, and domain adaptation. At Apollo.io, he fine-tuned a LLaMA model for the GTM AI Assistant, building the dataset in-house and validating output quality against a golden dataset. As co-founder of Stybe, he applied AI-driven recommendations to a consumer fashion product.
At SpotDraft, Pranav led three parallel workstreams: designing the routing, planning, and agent architecture for a multi-agent legal AI system; building an evaluation harness from scratch that other product teams could use to surface and track real customer failures; and constructing a large-scale data ingestion pipeline processing millions of documents with LLM-based metadata extraction. He also built an automated prompt optimization system that iterated on failing eval cases to improve agent performance without manual intervention.
Pranav is drawn to problems that sit at the intersection of model quality and system design, where getting the architecture right determines whether the product actually works in production. He is particularly interested in voice AI and the applied research challenges that come with building reliable, high-accuracy AI products at scale.
Pranav Pandey
Shipping features in AI is easy now. The real unlock is going deep enough on a use case that you can solve workflows end to end. That's what separates a product from a feature factory.”
Pranav Pandey
Pranav Pandey