2× GPU Utilization & 2,537× TPU Step Speedup
Led DCNv2 MoE ranking across ~750M–950M daily requests (44% → 98% GPU utilization; 42.4h → 15.6h 8× A100 training) and authored open-source TorchTPU Doctor diagnostics.
Hands-on technical and developer GTM advisory by Jigyasa Grover through Afterdays on production ML ranking, agentic AI orchestration, evaluation harnesses, and inference economics.
Through my independent practice, Afterdays, I work directly with founders, engineering VPs, and investors on architecture, evaluation, latency/cost trade-offs, and developer GTM. All advisory, design-partner, and expert-consultation work is conducted independently in my personal capacity, with selective bandwidth structured around focused 1:1 sessions and scoped architectural retainers.
Led DCNv2 MoE ranking across ~750M–950M daily requests (44% → 98% GPU utilization; 42.4h → 15.6h 8× A100 training) and authored open-source TorchTPU Doctor diagnostics.
Architected Bordo AI's LangGraph multi-LLM platform (Claude, Gemini, GPT) with RL dynamic tool routing and deterministic SQL ground-truth reward graders across 25+ organizations.
Advisor on Google Developer Advisory Board (gDAB), featured in Google I/O keynote (~150K views) and Gemini Omni Flash launch, plus developer GTM collaborations with Qodo, Parallel, and Fireflies.ai.
Includes startup advisory retainers, paid expert consultations, product design-partner feedback, and executive workshops delivered independently via Afterdays and ecosystem boards.
Separate workflows best served by state machines or single-step retrieval from genuine multi-agent topologies, auditing MCP schemas, memory, and routing.
Design calibrated multi-task ranking pipelines, real-time feature stores, cold-start smoothing, and sub-50ms serving architectures.
Build golden datasets, trajectory assertions, Pass@k benchmarks, calibrated LLM-as-a-Judge rubrics, and per-turn token circuit breakers.
Pressure-test API contracts, MCP/Agent Skill onboarding friction, benchmark-led marketing, and technical GTM positioning for AI startups selling to senior engineers and VPs.
Need to upskill your engineering org alongside architectural advisory? Pair an Afterdays retainer with private team workshops on Google ADK & MCP, coding-agent workflows, LLM evaluation, or ranking systems.
| Decision Factor | 1:1 Call (Intro · 5.0★) | Afterdays Advisory Retainer | Full-Time Staff ML Hire / Agency |
|---|---|---|---|
| Time to First Signal | Same-week 45–60 minute deep dive on architecture, agent evals, or developer GTM. | Biweekly or monthly design-partner reviews aligned to product milestones and launches. | 3–6+ month Staff ML recruiting cycle or multi-week agency discovery ramp. |
| Production Pattern Matching | Immediate reality check grounded in Uber, Bordo AI, Faire, Twitter/X, and Meta production ML systems. | Sustained guidance across multi-agent orchestration, ranking, Cloud TPU costs, and SDK onboarding. | Dedicated daily implementation bandwidth once recruited, onboarded, and ramped. |
| Ideal Trigger | Pre-launch sanity check, build-vs-buy agent decision, pricing/token FinOps audit, or VC technical diligence. | Seed to Series B AI startups needing Staff+ ML & developer GTM counsel alongside a lean core team. | When architecture is validated and you need full-time headcount to execute daily feature sprints. |
| Commitment & Flexibility | Single pay-per-session booking with zero lock-in. | Scoped fractional advisory engagement (independent personal capacity via Afterdays). | Full-time executive/Staff headcount budget or fixed multi-month consulting SOW. |
“Jigyasa is an excellent communicator, advisor and thought partner. She shared her journey authentically, making our conversation engaging and inspiring. She guided me toward a path tailored to my needs. She is a great thought partner and advisor. I would recommend her in a heartbeat.”
“Jigyasa was superb. She really helped me as a non-technical person begin to understand the finer details of building an LM for the new venture I'm working on. Highly recommended!”
“She is a true star, and yet is so down-to-earth and easy to talk to. She doesn't have any airs about her, and any question is a good question. I came out of the 30 mins meeting with lot of insights and resources that I can apply at my startup.”
“Her attention and suggestions significantly enhanced my deck for presenting to judges, mentors, and other Sac ecosystem stakeholders. She provided actionable advice that was insightful. I highly recommend her to any startup or innovator seeking expert guidance.”
Production ML ranking and personalization, feature stores, multi-agent orchestration vs. deterministic workflows, LLM evaluation harnesses, inference cost guardrails, and AI go-to-market (GTM) & developer positioning.
Yes. Through Afterdays (advising and consulting 30+ teams across Anthropic, Google gDAB, Windsurf, Base44, Okareo, and venture funds), she helps AI founders pressure-test technical positioning, SDK onboarding friction, benchmark-led marketing, and enterprise vs. PLG developer GTM.
Book focused 1:1 architecture or GTM strategy calls via Intro (5.0★), or inquire via Work With Me for ongoing startup advisory retainers through her independent practice, Afterdays.
A 1:1 Intro call or fractional Afterdays retainer delivers Staff+ production pattern matching (from Uber, Bordo AI, Faire, Twitter/X, and Meta) within days rather than waiting 3–6+ months for a full-time Bay Area Staff ML search or committing to a $150K+ agency SOW. Founders typically use advisory to de-risk agentic vs. deterministic architecture, evaluation harnesses, token economics, and developer GTM before or alongside hiring full-time engineers.