Petr Konstantinov

Principal Engineer · Applied AI & RAG · Embedded & Real-Time Video

hello@pitkonst.me +971-55-439-01-07 linkedin.com/in/petr-konstantinov Dubai, UAE (Resident - Open to Remote/Hybrid/Office/Relocation)

// Executive Summary

Principal engineer and systems architect with 15+ years building large-scale, latency-critical systems, now focused on production AI platforms: agentic architectures, grounded retrieval, and local LLM inference for environments where privacy and latency rule out hosted APIs. Co-founder and CTO of Jalinga, which I led from hardware prototype to a SaaS + hardware ecosystem deployed in 450+ interactive studios across MENA, Europe and the Americas, and through its AI transformation in 2024.

Expert in bridging hardware (embedded systems) and software (cloud, real-time video) - the span that carries an AI product from ingestion pipeline to GPU deployment without a handoff. Current work: a public MCP server over a 2000+ datasheet corpus, a hybrid semantic + lexical RAG backend that cites every claim, and an evaluation harness whose measurements, not intuition, decide which models and retrieval methods ship.

// Professional Experience

whatic.io Live

Founder & Principal AI Engineer
2026 — Present
An applied AI assistant for hardware engineers that recommends components backed by verifiable datasheet evidence. In production.
01 Architected and deployed a production agentic AI platform bridging LLMs with specialized engineering data, exposing a 2000+ datasheet corpus (200k+ indexed chunks) to external clients via a public Model Context Protocol (MCP) server.
02 Designed multi-agent workflows and a hybrid (semantic + lexical) RAG backend that grounds every claim in specific datasheet citations. Built a parametric extraction system achieving 94% precision and 88% coverage against a golden set without overfitting, indexing 160k+ parameters.
03 Engineered a rigorous, 23-experiment evaluation harness driving all model and retrieval decisions. Quantified the LLM-as-judge noise floor to establish a minimum detectable effect, preventing false-positive wins and reporting strict 95% confidence intervals (e.g., table classification at 92.7%±2.8%).
04 Built a 7-stage multimodal ingestion pipeline (layout + OCR + VLM) to convert dense, unstructured PDFs into structured vectors. Audited and resolved systemic data-loss defects in the indexing layer, recovering 70% of truncated table rows.
05 Optimized air-gapped local inference pipelines (vLLM, GGUF) on cost-effective GPUs, ensuring zero data exfiltration. Established robust CI/CD practices (4500+ automated tests, CI-gated CVE scanning).

Jalinga

CTO & Co-Founder
2015 — 2026
Global EdTech company developing interactive video studios and glass-board technology.

Leadership & Strategy

01 Co-founded the company and defined the technology roadmap, transforming a hardware prototype into a global SaaS + hardware business with 450+ studios deployed across multiple countries.
02 Built the R&D organization from a solo developer into a cross-functional team of 10+ senior engineers and product managers, fostering a culture of ownership and near-zero turnover.
03 Led technical expansion into MENA and EU markets, supporting enterprise and university customers.
04 Implemented structured Agile processes and automated QA, achieving predictable release cycles and consistent quarterly roadmap delivery.

Engineering & Architecture

05 Designed and delivered a proprietary GPU-accelerated real-time video compositing engine enabling natural “teacher + content” interaction with sub-second latency and zero dropped frames.
06 Architected and launched the cloud platform jalinga.cloud for content management and AI-assisted presentation creation, creating a recurring SaaS revenue stream and improving reliability for distributed customers.
07 Spearheaded the AI transformation of jalinga.cloud (2024), leading a cross-functional engineering team to architect and deploy production media-processing pipelines: automated video transcription, AI voice dubbing/translation, and LLM-driven generation of interactive presentations from raw text.
08 Built a WebRTC-based virtual classroom platform with a service mesh architecture, supporting large-scale concurrent sessions and resolving critical load and scalability issues.

TablePad

Co-Founder & Tech Lead
2013 — 2015
E-menu system for restaurants.
01 Architected and brought to market an enterprise B2B self-service platform and interactive digital menu system for the hospitality sector.
02 Designed a distributed architecture comprising Android client applications, a local coordination server, and a custom integration plugin for seamless bidirectional syncing with POS systems (iiko, MICROS).
03 Led the full SDLC from initial hardware/software concept to commercial release, deploying the solution into live, high-traffic restaurant environments.

Movicom

Head of Electronics Dept. / Lead Engineer
2008 — 2013
Robotic cameras for broadcast.
01 Developed hardware and firmware for stabilized camera heads and Robycam cable camera systems used at major sports events (Olympic Games, FIFA World Cup) under strict broadcast reliability and safety requirements.
02 Managed a team of electronics engineers, overseeing PCB design, embedded firmware (C/C++), and multi-axis motor control algorithms for robotic camera systems.
03 Delivered high-reliability platforms operating in harsh outdoor environments with millisecond-precision control, supporting numerous live broadcasts without critical failures.

// Core Competencies

Leadership & Strategy

R&D strategy, hands-on architecture, team management, Agile, SDLC, hiring & mentoring, stakeholder management.

Architecture & Cloud

Event-driven architecture, service mesh, REST API design, AWS, DigitalOcean, Terraform, Ansible, Docker, GitLab CI/CD, Postgres/pgvector, Prometheus/Grafana.

AI & LLM Engineering

Agentic orchestration & tool-calling, Model Context Protocol (MCP), retrieval-augmented generation (vector semantic search), hybrid retrieval (semantic + lexical), local LLM inference & optimization (vLLM, GGUF quantization), multimodal document extraction & ingestion, experiment-driven evaluation & benchmarking, open-weight models.

Video & Hardware

WebRTC, GPU acceleration (Direct 3D/NVENC), real-time compositing, embedded systems, touch-interface drivers.

Backend, Frontend & Security

Python, C#, TypeScript, C, .NET.
Security & Compliance: Data privacy, security reviews & audits.

// Education

Bauman Moscow State Technical University (BMSTU)

Master's Degree in Electronics Engineering & Technology
Focus: Robotics, Embedded Systems

// Languages

English Advanced (C1)
Russian Native