Resume
Kuldeep Pisda — engineering leader and former startup CTO with 6+ years building production AI systems: multi-LLM platforms, RAG and semantic search, voice/SMS automation, and platform architecture.
Engineering Leader · AI Systems · Platform Architecture
Bengaluru, India · +91 798 716 2852 · kdpisda@gmail.com · GitHub · LinkedIn · Blog
Summary
Engineering leader and former startup CTO with 6+ years building production AI systems end-to-end. Architect and primary author of multi-LLM platforms — agentic voice/SMS automation, RAG and semantic-search pipelines, and AI due-diligence engines built on Claude, GPT-4o, Gemini, and Perplexity. Has founded engineering functions, owned architecture, SRE, and security (SOC 2), and led small teams while staying deeply hands-on across backend, data, and infrastructure. Scaled multi-tenant platforms to tens of thousands of users; frequent international speaker (DjangoCon US & Europe, EuroPython).
Core skills
- Leadership — Engineering ownership & technical strategy, fractional CTO, team leadership & mentoring, hiring & engineering standards, TDD and code review.
- AI / ML — Agentic LLM systems (Claude, GPT-4o, Gemini, Perplexity), RAG & semantic search (Dify, pgvector, Elasticsearch), speech-to-text (Deepgram), prompt pipelines, voice/SMS agents.
- Architecture — Multi-tenant SaaS, event-driven systems (Inngest, Celery/RabbitMQ), “one backend, many clients”, API & data-sync platforms, payments & SSO (Keycloak).
- Infrastructure / SRE — AWS (ECS, S3, CloudFront), Vercel, Terraform IaC, Docker, GPU pipelines, CI/CD, monitoring & incident response (Datadog, New Relic, Sentry, PagerDuty), SOC 2.
- Languages & frameworks — Python, TypeScript/JavaScript, C++; Django/DRF, FastAPI, Next.js, React, Vue; PostgreSQL/pgvector, Elasticsearch.
Experience
Avoca Labs · Product Engineer
Mar 2025 – Present
- Architected the core of Avoca’s ServiceTitan CRM data-sync engine — an event-driven, multi-tenant pipeline (Inngest) mirroring 20+ entity types into Postgres with per-tenant concurrency and rate-throttling (global and per-team limits), incremental re-pull windows, append-only batch history, and auto-disable-on-failure for self-healing syncs at scale.
- Built the outbound call/SMS campaign engine as stateful, fan-out workflows orchestrating AI voice/SMS outreach: audience segmentation, availability-aware scheduling, speed-to-lead routing, plus SMS conversation search and opt-out compliance across tenants.
- Engineered the AI transcript-processing pipeline (unified
parseTranscript): high-volume Deepgram nova-3 speech-to-text (multilingual, AssemblyAI fallback) feeding Claude-based summarization and call insights via the Vercel AI SDK, in an event-driven TypeScript/Next.js monorepo (Supabase/Postgres, ClickHouse, Inngest).
Tejgyan Foundation · Fractional CTO (Contract)
Aug 2023 – Mar 2025
- Engaged as Fractional CTO to architect and lead engineering for a multi-product platform spanning a video/OTT app, LMS, ERP, donations, and events. Defined a “one backend, many clients” architecture — a single Django/DRF backend serving 15+ web frontends and native Android apps — and led a ~6-person engineering team across ~30 services.
- Built the AI content-intelligence pipeline over a large multilingual discourse corpus: Deepgram speech-to-text generates documents that feed a Dify RAG knowledge base powering a conversational assistant, with pgvector semantic search (Indic/Hindi SBERT embeddings) tuned for 100k+ documents.
- Engineered the OTT media pipeline from scratch: FFmpeg HLS transcoding with CPU and GPU workers (CUDA, on-demand via Vast.ai), delivered through Wasabi S3 + CDN with signed URLs; web (Video.js/HLS.js) and native Android (Media3/ExoPlayer) players.
- Stood up platform foundations: Keycloak SSO/IAM, a multi-tenant admin monorepo, event-driven Celery/RabbitMQ workers (10+ queues), and payments via Razorpay and ICICI e-NACH e-mandates.
Astel Ventures · Senior Full-Stack Developer (Contract)
Jul 2024 – May 2025
- Designed and built (primary author) a multi-LLM AI due-diligence engine: a 27-step pitch-deck analysis pipeline plus a citation-grounded deal analyser that generate comprehensive, structured investment reports, orchestrating GPT-4o, Claude Sonnet 4.5, Gemini 2.5, and Perplexity through a unified client layer with retry/backoff and per-tenant model selection.
- Engineered investor discovery & matching on Elasticsearch (custom edge-ngram + geo index over ~60k contact records) with Apollo.io enrichment pipelines and per-startup match scoring (Celery/RabbitMQ).
- Delivered it on a multi-tenant / whitelabel Django architecture (DRF, PostGIS, Vue 3, Gmail extension) with GitHub Actions, GHCR, and Trivy CI/CD.
CognifAI Health · Co-Founder & CTO
Aug 2023 – Jul 2024
- Co-founded the company and owned engineering end-to-end, setting technical direction and building the founding stack for a healthcare OTT/LMS platform that scaled to 20,000+ users.
- Built video/audio transcoding (FFmpeg, later AWS MediaConvert), payments (CCAvenue, Razorpay), and Keycloak auth/roles; stood up production AWS (ECS, S3, CloudFront) infrastructure with GitHub Actions CI/CD.
- Unified scattered legacy databases into a single PostgreSQL schema via Apache Airflow, and turned Django Admin into a dynamic CRM/ERP with custom IAM, PDF receipts, and real-time notifications.
Goldcast · Founding Engineer & Founding SRE
May 2021 – Mar 2023
- Founding engineer for a high-traffic live-events platform: led backend (Django/DRF) architecture and ran Locust load testing to define a data-driven scaling strategy.
- As founding SRE, designed a secure AWS architecture, migrated infrastructure to Terraform IaC, and built end-to-end monitoring/alerting from scratch (New Relic, CloudWatch, PagerDuty) with incident-response SOPs that reduced MTTR.
- Drove SOC 2 Type 1 & 2 compliance and penetration-test remediation (WAF, STS/CSP headers, A rating on securityheaders.com); mentored engineers and championed TDD, reducing production bugs.
CodeVector Labs · Core Team Member & Full-Stack Developer
Oct 2019 – Apr 2021
- Built e-Paarvai, a real-time cataract-detection platform serving ML image-classification models (Django, Celery, RabbitMQ) with a React dashboard used by Tamil Nadu state health officials (NHM); plus real-time products including a Node.js WebSocket multiplayer server and a pet-care CRM with PDF/Excel reporting.
Open source
- django-rls — PostgreSQL Row-Level Security for Django: declarative, model-level policies that enforce tenant isolation at the database layer; published on PyPI (BSD-3).
- KRONOS — research-grade plane-wave Density Functional Theory (DFT) engine in modern C++ (C++20) with CUDA/HIP/Metal GPU backends, validated against Quantum ESPRESSO to micro-Rydberg accuracy.
Publication
- Deep Networks for Brachial Plexus Nerves Segmentation and Detection Using Ultrasound Images — Pisda, K., Jain, P. & Sisodia, D. S. · Intl. Conf. on Information Systems & Management Science (ISMS), Springer LNNS vol. 303 (2022).
Education
- NIT Raipur — B.Tech, Computer Science & Engineering (7.47/10).
- IISc Bengaluru (CCE) — Introduction to High-Performance Computing (HPC) & Parallel Programming (short-term course) · Jun 2024 – Dec 2024.
Speaking
DjangoCon US 2022 · 2023 · 2024 · DjangoCon Europe 2026 · EuroPython 2023 · 2026 · SQL Saturday Silicon Valley 2023 · Google Cloud Community Days 2023 · APIDays India 2022.