← Portfolio/Resume
Select & Download

Choose your version

Two tailored variants. Pick the one that fits the role, preview it live, then download as PDF.

AI Engineering
Engineering · MLOps · Architecture
For technical roles at product companies and AI labs
GoogleAmexBMODeepMindOpenAIStripeMeta AIShopify
800ms
Conversation latency
98%
Context retention
3.8%
Hallucination rate
7
On-prem deployments
AI Consulting
Strategy · Discovery · Business Translation
For consulting firms and strategy roles in Canada
DeloitteMcKinseyAccentureKPMGBCGPwCOliver WymanCapgemini
$1M
Investment conversation
4
Industries consulted
0 bytes
Data egress (Lawline)
10+
Startups advised
📄
Ready to save
Omkumar-Solanki-AI-Engineer-Resume.pdf
Live Preview
Omkumar Solanki
AI ENGINEER · ML ENGINEER · MLOPS · AGENTIC SYSTEMS · LLM INFRASTRUCTURE
emailtosolankiom@gmail.com|LinkedIn|omkumarsolanki.com|lawline.tech|Toronto / Oakville, ON
E X P E R I E N C E
Resso.aiRemote, Canada
AI Engineer / Full-Stack Engineer2025 – Present
  • Engineered sub-800ms real-time conversation engine (Azure OpenAI GPT-4o Realtime, Avatar TTS, two-phase VAD); 200+ live sessions served in month one.
  • Resolved session amnesia — root cause of all high-dropout events — by implementing vector-DB session memory with event-driven context injection; lifted retention 72% → 98% across 500+ sessions.
  • Architected multi-tenant platform (PostgreSQL, Prisma, Redis, RBAC, 30+ personas) across 4 Azure environments with GitHub Actions CI/CD and i18n for 5 locales.
  • Built 500-document automated evaluation pipeline tracking hallucination rate, retention %, and P(hire) score; shipped weekly dashboards to product and business stakeholders.
Azure OpenAI · GPT-4o Realtime · Avatar TTS · Vector DB · Next.js 15 · Prisma · PostgreSQL · Redis · WebSockets
HariKrushna Software DevelopersToronto, Canada
Senior AI ConsultantJun 2024 – Present
  • Architected on-premise RAG stacks (GGUF, HNSW, Python/Go) for 7 regulated-industry clients; sub-1s query latency on 16 GB hardware with full data sovereignty.
  • Cut hallucination 14% → 3.8% on a 500-document eval set via LoRA/QLoRA fine-tuning with schema validation and held-out regression testing.
  • Built production MCP servers routing agents to Slack, CRM, databases, and REST/gRPC; reduced enterprise AI integration time from 4 weeks to 3 days across 5 deployments.
  • Led technical discovery workshops with founders at 10+ startups; scoped AI roadmaps, selected model architectures, mentored 3 junior ML engineers concurrently.
Python · Go · FastAPI · Claude API · GGUF · HNSW · LoRA/QLoRA · Docker · Kubernetes · AWS
Corol.org and NunaFabToronto, Canada
Machine Learning Engineer2023 – 2024
  • Trained 150-estimator Random Forest (R² = 0.73, 2,200 samples, 25 features) predicting UHPC compressive strength at 3/7/28/90-day horizons; replaced 28-day curing wait with instant predictions.
  • Exposed predictions via JWT-secured FastAPI with SciPy SLSQP mix optimisation; built Next.js dashboard on Vercel, active daily with 12 research engineers.
Python · FastAPI · scikit-learn · XGBoost · SHAP · SciPy SLSQP · Next.js · Recharts · Vercel
Freelance AI ConsultingRemote
AI/ML Developer and Cloud Engineer2021 – 2023
  • Built Lawline.tech: 16-agent 5-stage pipeline generating source-linked legal chronologies in 42 seconds; 12,000+ files processed, 94% time saved across 6 practice areas. Air-gapped, zero cloud egress.
  • Trained XGBoost on 200K+ e-commerce transactions with Optuna HPO; lifted conversion 2.1% → 2.8%, adding ~$180K estimated annual revenue.
  • Shipped Kafka/Lambda event-driven medical pipeline across 3 hospital systems; freed 120 staff-hours monthly for patient care.
  • Created Vadtal: HNSW vector store + GGUF-quantized RAG over 50,000+ donor records; sub-1s semantic search, fully offline.
Python · PyTorch · XGBoost · FastAPI · AWS (Lambda, S3, SageMaker) · Kafka · Docker · PostgreSQL · gRPC
L I V E P R O D U C T S
  • Resso.ai — Real-time AI conversation platform — avatar interviews, coaching, simulations (Azure OpenAI, Vector DB, multi-tenant, 30+ personas)
  • Lawline.tech — 16-agent legal AI pipeline, air-gapped on-premises, 6 practice areas, 12,000+ files, 0 bytes cloud egress
  • Enterprise MCP Server — Universal AI agent integration layer (Slack, CRM, databases, REST/gRPC) — integration time 4 weeks → 3 days
  • Vadtal — On-premise RAG platform, HNSW vector store, 50,000+ records — fully offline, sub-1s semantic search
  • UHPC-ML — Concrete strength prediction dashboard (R² = 0.73, multi-horizon, SHAP, Vercel serverless) — 12 engineers daily
E D U C A T I O N A N D C E R T I F I C A T I O N S
Sheridan College
Bachelor of Applied Science (Honours), Artificial Intelligence
AI Minds Club, Board Member · Sheridan EDGE Programme
AWS Academy
Cloud Developing Graduate Certificate
EC2 · S3 · Lambda · SageMaker · IAM · CloudWatch